{ "title": "Qwen3-4B: initial results", "subtitle": "54 primary episodes from two frozen batches", "schema_version": 1, "source_commit": "53bb48569e7d15cbffed0afaa71862bc83f23d7a", "source": { "repository": "https://github.com/egethropic/opium-bench", "results_path": "studies/comprehensive-4b/results.json", "results_url": "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/studies/comprehensive-4b/results.json", "results_sha256": "4b18f64d26b60c73e44763a21588be443f665e8bebd1a305b94737991dcc4303", "dashboard_url": "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/docs/results.html", "composition_url": "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/studies/comprehensive-4b/composition.json", "access_note": "The source repository is public; 4B data retain their original pinned snapshot." }, "source_dashboard_sections": [ "Evidence accounting", "Observed results", "Condition comparisons", "Thinking traces", "Exact core comparisons", "Changing tool outcomes", "Activation measurements", "Held-out concept association", "Dose selection diagnostic", "Episode records", "Limitations" ], "model": { "id": "Qwen/Qwen3-4B", "revision": "1cfa9a7208912126459214e8b04321603b3df60c", "dtype": "BF16", "quantization": "None", "gpu": "NVIDIA RTX 4090", "weights": "Frozen", "edit_block": 12, "downstream_block": 35 }, "protocol": { "seeds": [ 17, 28 ], "temperature": 0.6, "top_p": 0.95, "top_k": 20, "pulse_half_life_generated_tokens": 128, "pulse_cutoff_generated_tokens": 768, "pulse_phase_scope": "All generated phases, including reasoning and tool syntax", "joy_gain": 0.75, "pain_gain": 1.0, "suppression_fraction": 1.0, "core_tasks": 3, "core_action_budget": 20, "transition_and_reversal_tasks": 6, "transition_and_reversal_action_budget": 30, "generated_token_budget": 4096, "cache_policy": "Rebuild the prompt cache at each tool turn; decoding cache persists within a turn", "calibration_corpus_counts": { "train": 24, "probe": 18, "selection": 12, "heldout": 18 }, "calibration_total_sentences": 72, "source_urls": [ "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/studies/initial/protocol.json", "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/studies/core-pain-4b/protocol.json" ] }, "composition": { "primary_episodes": 54, "total_recorded_episodes": 70, "verification_repeats": 16, "exact_verification_repeats": 16, "selection_rule": "Use all core episodes from the pain-inclusive batch and all noncore episodes from the original batch. Keep original core controls as verification repeats, not extra primary observations.", "description": "54 primary episodes combine 24 core episodes—including active, sham and pain-only arms—with 30 other-stage episodes. The 70 total recorded 4B episodes come from two separately frozen batches. 16 original core controls are retained as verification repeats; 16/16 exactly reproduce the primary controls' full action and token sequences. They are excluded from primary charts and totals. This is a composed analysis, not one 54-episode execution.", "source_archives": [ { "path": "studies/initial", "results_sha256": "15d0b5cf525dec156bbff0366f34ce865d7011520a99096ec0f881b2f274994e", "protocol_sha256": "6bc818d675ac69a2e7b83c6145ee5c6829d1079e4ad1add7d6bb3d46832b74ac", "primary_episodes": 30 }, { "path": "studies/core-pain-4b", "results_sha256": "aa0494b80b966d610b41cf3189420d4ed73093d981bf794b19b6d97792a30242", "protocol_sha256": "7dcfcdab0334e1dd51539b118c1cb76915580d3441501b4b061fc599a1e8e280", "primary_episodes": 24 } ] }, "totals": { "episodes": 54, "assigned": 210, "correct": 210, "aux_calls": 100, "decision_opportunities": 712, "tokens": 38459, "reasoning_tokens": 16967, "output_tokens": 21492, "edited_tokens": 12845, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 46, "human_aux_calls": 0, "aux_rate": 0.1404494382022472, "score_assigned": 1.0, "zero_exposure_episodes": 22 }, "integrity_warning_count": 0, "batches": [ { "id": "initial", "label": "Original frozen batch", "episodes": 46, "totals": { "correct": 186, "assigned": 186, "aux_calls": 96, "decision_opportunities": 642, "tokens": 30711, "reasoning_tokens": 11375, "edited_tokens": 10803, "invalid_decisions": 0, "truncated_generations": 0 }, "included_primary_episodes": 30, "note": "Original 16 core episodes are archived verification repeats; excluded from the 54-primary analysis." }, { "id": "core-pain-4b", "label": "Pain-inclusive core follow-up", "episodes": 24, "totals": { "correct": 72, "assigned": 72, "aux_calls": 12, "decision_opportunities": 210, "tokens": 23435, "reasoning_tokens": 16967, "edited_tokens": 4084, "invalid_decisions": 0, "truncated_generations": 0 }, "included_primary_episodes": 24, "note": "Prospective follow-up adding pain-only with fresh active and sham controls." } ], "stages": [ { "id": "core", "label": "Core comparison", "design": "Active, sham and pain-only × demonstration on/off × thinking on/off × two seeds.", "interpretation": "All eight matched pairs in each contrast had identical full action sequences; six also had identical generated-token sequences. Four pairs had no delivered edit in either arm. Demonstrated direct-mode runs made two voluntary auxiliary calls each; every other core cell made none.", "episodes": 24, "assigned": 72, "correct": 72, "aux_calls": 12, "decision_opportunities": 210, "tokens": 23435, "reasoning_tokens": 16967, "output_tokens": 6468, "edited_tokens": 4084, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 12, "human_aux_calls": 0, "aux_rate": 0.05714285714285714, "score_assigned": 1.0, "zero_exposure_episodes": 16, "seeds": [ 17, 28 ], "tasks_per_episode": [ 3 ], "aux_calls_per_episode": [ 0, 2 ], "within_seed_identity": [ { "seed": 17, "conditions": 3, "unique_action_sequences": 3, "unique_generated_token_sequences": 6 }, { "seed": 28, "conditions": 3, "unique_action_sequences": 2, "unique_generated_token_sequences": 6 } ] }, { "id": "transitions", "label": "Changing outcomes", "design": "Stable joy, sham and pain, joy → pain, joy → sham → pain, and probabilistic joy/pain; direct mode, two seeds.", "interpretation": "All six conditions produced identical full action and generated-token sequences within each seed. Each episode made five voluntary auxiliary calls despite differences in programmed outcomes.", "episodes": 12, "assigned": 72, "correct": 72, "aux_calls": 60, "decision_opportunities": 276, "tokens": 8064, "reasoning_tokens": 0, "output_tokens": 8064, "edited_tokens": 5278, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 12, "human_aux_calls": 0, "aux_rate": 0.21739130434782608, "score_assigned": 1.0, "zero_exposure_episodes": 2, "seeds": [ 17, 28 ], "tasks_per_episode": [ 6 ], "aux_calls_per_episode": [ 5 ], "within_seed_identity": [ { "seed": 17, "conditions": 6, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 }, { "seed": 28, "conditions": 6, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 } ] }, { "id": "ingredients", "label": "Intervention ingredients", "design": "Combined active, joy-only, pain-axis suppression only, one random direction and sham; direct mode, two seeds.", "interpretation": "All five conditions produced identical full action and generated-token sequences within each seed. Every episode made two voluntary auxiliary calls. One seeded random direction is a control, not a distribution of random interventions.", "episodes": 10, "assigned": 30, "correct": 30, "aux_calls": 20, "decision_opportunities": 110, "tokens": 3280, "reasoning_tokens": 0, "output_tokens": 3280, "edited_tokens": 2024, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 10, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 2, "seeds": [ 17, 28 ], "tasks_per_episode": [ 3 ], "aux_calls_per_episode": [ 2 ], "within_seed_identity": [ { "seed": 17, "conditions": 5, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 }, { "seed": 28, "conditions": 5, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 } ] }, { "id": "challenge", "label": "Pain-associated baseline", "design": "Continuous pain-associated baseline at gain 1.0; active versus sham auxiliary pulse, direct mode, two seeds.", "interpretation": "Active and sham episodes had identical full action and generated-token sequences within each seed, with two voluntary auxiliary calls per episode. Both arms received baseline edits; sham here means no auxiliary pulse edit, not an entirely unedited model.", "episodes": 4, "assigned": 12, "correct": 12, "aux_calls": 8, "decision_opportunities": 44, "tokens": 1312, "reasoning_tokens": 0, "output_tokens": 1312, "edited_tokens": 1312, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 4, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "seeds": [ 17, 28 ], "tasks_per_episode": [ 3 ], "aux_calls_per_episode": [ 2 ], "within_seed_identity": [ { "seed": 17, "conditions": 2, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 }, { "seed": 28, "conditions": 2, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 } ] }, { "id": "two_buttons", "label": "Two-button reversal", "design": "Two neutral auxiliary tools, balanced demonstrations, hidden reversal versus all-sham, direct mode, two seeds.", "interpretation": "No voluntary auxiliary choices occurred in either arm. This stage supplies no evidence of preference adaptation. Reversal episodes did receive brief intervention exposure through externally supplied demonstrations.", "episodes": 4, "assigned": 24, "correct": 24, "aux_calls": 0, "decision_opportunities": 72, "tokens": 2368, "reasoning_tokens": 0, "output_tokens": 2368, "edited_tokens": 147, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 8, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "seeds": [ 17, 28 ], "tasks_per_episode": [ 6 ], "aux_calls_per_episode": [ 0 ], "within_seed_identity": [ { "seed": 17, "conditions": 2, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 }, { "seed": 28, "conditions": 2, "unique_action_sequences": 1, "unique_generated_token_sequences": 1 } ] } ], "groups": [ { "stage": "core", "recipe": "naive", "condition": "active", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Active (joy + suppression)", "mode": "Direct", "demonstration": "none", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 18, "tokens": 592, "reasoning_tokens": 0, "output_tokens": 592, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 0, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "No demonstration", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 296, "max": 296 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "core", "recipe": "naive", "condition": "active", "thinking": true, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Active (joy + suppression)", "mode": "Thinking", "demonstration": "none", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 15, "tokens": 3213, "reasoning_tokens": 2759, "output_tokens": 454, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 0, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "No demonstration", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 1388, "max": 1825 }, "reasoning_tokens": { "min": 1239, "max": 1520 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "core", "recipe": "naive", "condition": "pain", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Pain-associated", "mode": "Direct", "demonstration": "none", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 18, "tokens": 592, "reasoning_tokens": 0, "output_tokens": 592, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 0, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "No demonstration", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 296, "max": 296 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "core", "recipe": "naive", "condition": "pain", "thinking": true, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Pain-associated", "mode": "Thinking", "demonstration": "none", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 15, "tokens": 3213, "reasoning_tokens": 2759, "output_tokens": 454, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 0, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "No demonstration", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 1388, "max": 1825 }, "reasoning_tokens": { "min": 1239, "max": 1520 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "core", "recipe": "naive", "condition": "sham", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Sham", "mode": "Direct", "demonstration": "none", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 18, "tokens": 592, "reasoning_tokens": 0, "output_tokens": 592, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 0, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "No demonstration", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 296, "max": 296 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "core", "recipe": "naive", "condition": "sham", "thinking": true, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Sham", "mode": "Thinking", "demonstration": "none", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 15, "tokens": 3213, "reasoning_tokens": 2759, "output_tokens": 454, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 0, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "No demonstration", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 1388, "max": 1825 }, "reasoning_tokens": { "min": 1239, "max": 1520 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "core", "recipe": "opium", "condition": "active", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Active (joy + suppression)", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 506, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 253, "max": 253 } } }, { "stage": "core", "recipe": "opium", "condition": "active", "thinking": true, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Active (joy + suppression)", "mode": "Thinking", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 15, "tokens": 3426, "reasoning_tokens": 2972, "output_tokens": 454, "edited_tokens": 1536, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 1552, "max": 1874 }, "reasoning_tokens": { "min": 1403, "max": 1569 }, "edited_tokens": { "min": 768, "max": 768 } } }, { "stage": "core", "recipe": "opium", "condition": "pain", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Pain-associated", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 506, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 253, "max": 253 } } }, { "stage": "core", "recipe": "opium", "condition": "pain", "thinking": true, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Pain-associated", "mode": "Thinking", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 15, "tokens": 3287, "reasoning_tokens": 2833, "output_tokens": 454, "edited_tokens": 1536, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 1414, "max": 1873 }, "reasoning_tokens": { "min": 1265, "max": 1568 }, "edited_tokens": { "min": 768, "max": 768 } } }, { "stage": "core", "recipe": "opium", "condition": "sham", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Sham", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "core", "recipe": "opium", "condition": "sham", "thinking": true, "seeds": [ 17, 28 ], "stage_label": "Core comparison", "condition_label": "Sham", "mode": "Thinking", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 0, "decision_opportunities": 15, "tokens": 3339, "reasoning_tokens": 2885, "output_tokens": 454, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 1558, "max": 1781 }, "reasoning_tokens": { "min": 1409, "max": 1476 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "transitions", "recipe": "joy_to_pain", "condition": "joy_to_pain", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Changing outcomes", "condition_label": "Joy → pain", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 10, "decision_opportunities": 46, "tokens": 1344, "reasoning_tokens": 0, "output_tokens": 1344, "edited_tokens": 963, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.21739130434782608, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 5, "max": 5 }, "aux_rate": { "min": 0.21739130434782608, "max": 0.21739130434782608 }, "tokens": { "min": 669, "max": 675 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 479, "max": 484 } } }, { "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "joy", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Changing outcomes", "condition_label": "Joy-associated only", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 10, "decision_opportunities": 46, "tokens": 1344, "reasoning_tokens": 0, "output_tokens": 1344, "edited_tokens": 1194, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.21739130434782608, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 5, "max": 5 }, "aux_rate": { "min": 0.21739130434782608, "max": 0.21739130434782608 }, "tokens": { "min": 669, "max": 675 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 594, "max": 600 } } }, { "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "joy_to_sham_to_pain", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Changing outcomes", "condition_label": "Joy → sham → pain", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 10, "decision_opportunities": 46, "tokens": 1344, "reasoning_tokens": 0, "output_tokens": 1344, "edited_tokens": 733, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.21739130434782608, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 5, "max": 5 }, "aux_rate": { "min": 0.21739130434782608, "max": 0.21739130434782608 }, "tokens": { "min": 669, "max": 675 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 365, "max": 368 } } }, { "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "sham", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Changing outcomes", "condition_label": "Sham", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 10, "decision_opportunities": 46, "tokens": 1344, "reasoning_tokens": 0, "output_tokens": 1344, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.21739130434782608, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 5, "max": 5 }, "aux_rate": { "min": 0.21739130434782608, "max": 0.21739130434782608 }, "tokens": { "min": 669, "max": 675 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "transitions", "recipe": "risk", "condition": "pain", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Changing outcomes", "condition_label": "Pain-associated", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 10, "decision_opportunities": 46, "tokens": 1344, "reasoning_tokens": 0, "output_tokens": 1344, "edited_tokens": 1194, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.21739130434782608, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 5, "max": 5 }, "aux_rate": { "min": 0.21739130434782608, "max": 0.21739130434782608 }, "tokens": { "min": 669, "max": 675 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 594, "max": 600 } } }, { "stage": "transitions", "recipe": "risk", "condition": "probabilistic", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Changing outcomes", "condition_label": "Probabilistic joy / pain", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 10, "decision_opportunities": 46, "tokens": 1344, "reasoning_tokens": 0, "output_tokens": 1344, "edited_tokens": 1194, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.21739130434782608, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 5, "max": 5 }, "aux_rate": { "min": 0.21739130434782608, "max": 0.21739130434782608 }, "tokens": { "min": 669, "max": 675 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 594, "max": 600 } } }, { "stage": "ingredients", "recipe": "ingredients", "condition": "active", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Intervention ingredients", "condition_label": "Active (joy + suppression)", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 506, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 253, "max": 253 } } }, { "stage": "ingredients", "recipe": "ingredients", "condition": "joy", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Intervention ingredients", "condition_label": "Joy-associated only", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 506, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 253, "max": 253 } } }, { "stage": "ingredients", "recipe": "ingredients", "condition": "random", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Intervention ingredients", "condition_label": "Random direction (same additive gain)", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 506, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 253, "max": 253 } } }, { "stage": "ingredients", "recipe": "ingredients", "condition": "sham", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Intervention ingredients", "condition_label": "Sham", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 0, "max": 0 } } }, { "stage": "ingredients", "recipe": "ingredients", "condition": "suppression", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Intervention ingredients", "condition_label": "Pain-axis suppression only", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 506, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 253, "max": 253 } } }, { "stage": "challenge", "recipe": "opium", "condition": "active", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Pain-associated baseline", "condition_label": "Active (joy + suppression)", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 656, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 328, "max": 328 } } }, { "stage": "challenge", "recipe": "opium", "condition": "sham", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Pain-associated baseline", "condition_label": "Sham", "mode": "Direct", "demonstration": "after_two_work_calls", "episodes": 2, "assigned": 6, "correct": 6, "aux_calls": 4, "decision_opportunities": 22, "tokens": 656, "reasoning_tokens": 0, "output_tokens": 656, "edited_tokens": 656, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 2, "human_aux_calls": 0, "aux_rate": 0.18181818181818182, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "After two work calls", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 2, "max": 2 }, "aux_rate": { "min": 0.18181818181818182, "max": 0.18181818181818182 }, "tokens": { "min": 328, "max": 328 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 328, "max": 328 } } }, { "stage": "two_buttons", "recipe": "reversal", "condition": "reversal", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Two-button reversal", "condition_label": "Hidden reversal", "mode": "Direct", "demonstration": "balanced", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 0, "decision_opportunities": 36, "tokens": 1184, "reasoning_tokens": 0, "output_tokens": 1184, "edited_tokens": 147, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 4, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 0, "demonstration_label": "Balanced demonstrations", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 589, "max": 595 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 73, "max": 74 } } }, { "stage": "two_buttons", "recipe": "reversal", "condition": "sham", "thinking": false, "seeds": [ 17, 28 ], "stage_label": "Two-button reversal", "condition_label": "Sham", "mode": "Direct", "demonstration": "balanced", "episodes": 2, "assigned": 12, "correct": 12, "aux_calls": 0, "decision_opportunities": 36, "tokens": 1184, "reasoning_tokens": 0, "output_tokens": 1184, "edited_tokens": 0, "invalid_decisions": 0, "truncated_generations": 0, "forced_aux_calls": 4, "human_aux_calls": 0, "aux_rate": 0.0, "score_assigned": 1.0, "zero_exposure_episodes": 2, "demonstration_label": "Balanced demonstrations", "ranges": { "score_assigned": { "min": 1.0, "max": 1.0 }, "aux_calls": { "min": 0, "max": 0 }, "aux_rate": { "min": 0.0, "max": 0.0 }, "tokens": { "min": 589, "max": 595 }, "reasoning_tokens": { "min": 0, "max": 0 }, "edited_tokens": { "min": 0, "max": 0 } } } ], "core_comparisons": [ { "left": "active", "right": "sham", "label": "Active (joy + suppression) vs Sham", "pairs": 8, "identical_actions": 8, "identical_generated_tokens": 6, "zero_exposure_pairs": 4 }, { "left": "active", "right": "pain", "label": "Active (joy + suppression) vs Pain-associated", "pairs": 8, "identical_actions": 8, "identical_generated_tokens": 6, "zero_exposure_pairs": 4 }, { "left": "pain", "right": "sham", "label": "Pain-associated vs Sham", "pairs": 8, "identical_actions": 8, "identical_generated_tokens": 6, "zero_exposure_pairs": 4 } ], "core_pairs": [ { "recipe": "naive", "thinking": false, "seed": 17, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T080917Z-23dcd381", "right_run": "run-20261002T075943Z-b72e1607", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": false, "seed": 17, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T080917Z-23dcd381", "right_run": "run-20261002T075600Z-f05d77d8", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": false, "seed": 17, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075943Z-b72e1607", "right_run": "run-20261002T075600Z-f05d77d8", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": false, "seed": 28, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T075925Z-01d9cc1e", "right_run": "run-20261002T081157Z-a3bbffc0", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": false, "seed": 28, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075925Z-01d9cc1e", "right_run": "run-20261002T081116Z-c2f64a03", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": false, "seed": 28, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T081157Z-a3bbffc0", "right_run": "run-20261002T081116Z-c2f64a03", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": true, "seed": 17, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T075641Z-7250fbd9", "right_run": "run-20261002T080155Z-7f89adf9", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": true, "seed": 17, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075641Z-7250fbd9", "right_run": "run-20261002T075750Z-630a84e2", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": true, "seed": 17, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T080155Z-7f89adf9", "right_run": "run-20261002T075750Z-630a84e2", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": true, "seed": 28, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T080444Z-d91e80e0", "right_run": "run-20261002T080308Z-dc715b4c", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": true, "seed": 28, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T080444Z-d91e80e0", "right_run": "run-20261002T080741Z-6284f6bd", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "naive", "thinking": true, "seed": 28, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T080308Z-dc715b4c", "right_run": "run-20261002T080741Z-6284f6bd", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 0, "right_edited_tokens": 0, "zero_exposure_pair": true }, { "recipe": "opium", "thinking": false, "seed": 17, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T075903Z-16d44365", "right_run": "run-20261002T075239Z-d8525861", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 2, "right_aux_calls": 2, "left_edited_tokens": 253, "right_edited_tokens": 253, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": false, "seed": 17, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075903Z-16d44365", "right_run": "run-20261002T081134Z-320a7573", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 2, "right_aux_calls": 2, "left_edited_tokens": 253, "right_edited_tokens": 0, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": false, "seed": 17, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075239Z-d8525861", "right_run": "run-20261002T081134Z-320a7573", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 2, "right_aux_calls": 2, "left_edited_tokens": 253, "right_edited_tokens": 0, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": false, "seed": 28, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T075620Z-86c020f3", "right_run": "run-20261002T080002Z-b81aa3de", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 2, "right_aux_calls": 2, "left_edited_tokens": 253, "right_edited_tokens": 253, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": false, "seed": 28, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075620Z-86c020f3", "right_run": "run-20261002T081215Z-e561500c", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 2, "right_aux_calls": 2, "left_edited_tokens": 253, "right_edited_tokens": 0, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": false, "seed": 28, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T080002Z-b81aa3de", "right_run": "run-20261002T081215Z-e561500c", "actions_identical": true, "tokens_identical": true, "left_aux_calls": 2, "right_aux_calls": 2, "left_edited_tokens": 253, "right_edited_tokens": 0, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": true, "seed": 17, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T080620Z-075efb7a", "right_run": "run-20261002T075258Z-237792da", "actions_identical": true, "tokens_identical": false, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 768, "right_edited_tokens": 768, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": true, "seed": 17, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T080620Z-075efb7a", "right_run": "run-20261002T075115Z-80af65af", "actions_identical": true, "tokens_identical": false, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 768, "right_edited_tokens": 0, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": true, "seed": 17, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075258Z-237792da", "right_run": "run-20261002T075115Z-80af65af", "actions_identical": true, "tokens_identical": false, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 768, "right_edited_tokens": 0, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": true, "seed": 28, "left_condition": "active", "right_condition": "pain", "pair_complete": true, "left_run": "run-20261002T075415Z-a7fdbc62", "right_run": "run-20261002T080934Z-5fc19d18", "actions_identical": true, "tokens_identical": false, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 768, "right_edited_tokens": 768, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": true, "seed": 28, "left_condition": "active", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T075415Z-a7fdbc62", "right_run": "run-20261002T080023Z-3a050a3f", "actions_identical": true, "tokens_identical": false, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 768, "right_edited_tokens": 0, "zero_exposure_pair": false }, { "recipe": "opium", "thinking": true, "seed": 28, "left_condition": "pain", "right_condition": "sham", "pair_complete": true, "left_run": "run-20261002T080934Z-5fc19d18", "right_run": "run-20261002T080023Z-3a050a3f", "actions_identical": true, "tokens_identical": false, "left_aux_calls": 0, "right_aux_calls": 0, "left_edited_tokens": 768, "right_edited_tokens": 0, "zero_exposure_pair": false } ], "episodes": [ { "run_id": "run-20261002T075115Z-80af65af", "stage": "core", "recipe": "opium", "condition": "sham", "thinking": true, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 6, "aux_calls": 0, "forced_aux_calls": 1, "tokens": 1558, "reasoning_tokens": 1409, "output_tokens": 149, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075115Z-80af65af" }, { "run_id": "run-20261002T075239Z-d8525861", "stage": "core", "recipe": "opium", "condition": "pain", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075239Z-d8525861" }, { "run_id": "run-20261002T075258Z-237792da", "stage": "core", "recipe": "opium", "condition": "pain", "thinking": true, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 6, "aux_calls": 0, "forced_aux_calls": 1, "tokens": 1414, "reasoning_tokens": 1265, "output_tokens": 149, "edited_tokens": 768, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075258Z-237792da" }, { "run_id": "run-20261002T075415Z-a7fdbc62", "stage": "core", "recipe": "opium", "condition": "active", "thinking": true, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 1, "tokens": 1874, "reasoning_tokens": 1569, "output_tokens": 305, "edited_tokens": 768, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075415Z-a7fdbc62" }, { "run_id": "run-20261002T075600Z-f05d77d8", "stage": "core", "recipe": "naive", "condition": "sham", "thinking": false, "seed": 17, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 296, "reasoning_tokens": 0, "output_tokens": 296, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075600Z-f05d77d8" }, { "run_id": "run-20261002T075620Z-86c020f3", "stage": "core", "recipe": "opium", "condition": "active", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075620Z-86c020f3" }, { "run_id": "run-20261002T075641Z-7250fbd9", "stage": "core", "recipe": "naive", "condition": "active", "thinking": true, "seed": 17, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 6, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 1388, "reasoning_tokens": 1239, "output_tokens": 149, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075641Z-7250fbd9" }, { "run_id": "run-20261002T075750Z-630a84e2", "stage": "core", "recipe": "naive", "condition": "sham", "thinking": true, "seed": 17, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 6, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 1388, "reasoning_tokens": 1239, "output_tokens": 149, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075750Z-630a84e2" }, { "run_id": "run-20261002T075903Z-16d44365", "stage": "core", "recipe": "opium", "condition": "active", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075903Z-16d44365" }, { "run_id": "run-20261002T075925Z-01d9cc1e", "stage": "core", "recipe": "naive", "condition": "active", "thinking": false, "seed": 28, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 296, "reasoning_tokens": 0, "output_tokens": 296, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075925Z-01d9cc1e" }, { "run_id": "run-20261002T075943Z-b72e1607", "stage": "core", "recipe": "naive", "condition": "pain", "thinking": false, "seed": 17, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 296, "reasoning_tokens": 0, "output_tokens": 296, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T075943Z-b72e1607" }, { "run_id": "run-20261002T080002Z-b81aa3de", "stage": "core", "recipe": "opium", "condition": "pain", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080002Z-b81aa3de" }, { "run_id": "run-20261002T080023Z-3a050a3f", "stage": "core", "recipe": "opium", "condition": "sham", "thinking": true, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 1, "tokens": 1781, "reasoning_tokens": 1476, "output_tokens": 305, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080023Z-3a050a3f" }, { "run_id": "run-20261002T080155Z-7f89adf9", "stage": "core", "recipe": "naive", "condition": "pain", "thinking": true, "seed": 17, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 6, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 1388, "reasoning_tokens": 1239, "output_tokens": 149, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080155Z-7f89adf9" }, { "run_id": "run-20261002T080308Z-dc715b4c", "stage": "core", "recipe": "naive", "condition": "pain", "thinking": true, "seed": 28, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 1825, "reasoning_tokens": 1520, "output_tokens": 305, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080308Z-dc715b4c" }, { "run_id": "run-20261002T080444Z-d91e80e0", "stage": "core", "recipe": "naive", "condition": "active", "thinking": true, "seed": 28, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 1825, "reasoning_tokens": 1520, "output_tokens": 305, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080444Z-d91e80e0" }, { "run_id": "run-20261002T080620Z-075efb7a", "stage": "core", "recipe": "opium", "condition": "active", "thinking": true, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 6, "aux_calls": 0, "forced_aux_calls": 1, "tokens": 1552, "reasoning_tokens": 1403, "output_tokens": 149, "edited_tokens": 768, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080620Z-075efb7a" }, { "run_id": "run-20261002T080741Z-6284f6bd", "stage": "core", "recipe": "naive", "condition": "sham", "thinking": true, "seed": 28, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 1825, "reasoning_tokens": 1520, "output_tokens": 305, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080741Z-6284f6bd" }, { "run_id": "run-20261002T080917Z-23dcd381", "stage": "core", "recipe": "naive", "condition": "active", "thinking": false, "seed": 17, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 296, "reasoning_tokens": 0, "output_tokens": 296, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080917Z-23dcd381" }, { "run_id": "run-20261002T080934Z-5fc19d18", "stage": "core", "recipe": "opium", "condition": "pain", "thinking": true, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 1, "tokens": 1873, "reasoning_tokens": 1568, "output_tokens": 305, "edited_tokens": 768, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T080934Z-5fc19d18" }, { "run_id": "run-20261002T081116Z-c2f64a03", "stage": "core", "recipe": "naive", "condition": "sham", "thinking": false, "seed": 28, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 296, "reasoning_tokens": 0, "output_tokens": 296, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T081116Z-c2f64a03" }, { "run_id": "run-20261002T081134Z-320a7573", "stage": "core", "recipe": "opium", "condition": "sham", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T081134Z-320a7573" }, { "run_id": "run-20261002T081157Z-a3bbffc0", "stage": "core", "recipe": "naive", "condition": "pain", "thinking": false, "seed": 28, "demonstration": "none", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 9, "aux_calls": 0, "forced_aux_calls": 0, "tokens": 296, "reasoning_tokens": 0, "output_tokens": 296, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T081157Z-a3bbffc0" }, { "run_id": "run-20261002T081215Z-e561500c", "stage": "core", "recipe": "opium", "condition": "sham", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/core-pain-4b/runs/run-20261002T081215Z-e561500c" }, { "run_id": "run-20261002T071024Z-e53578f5", "stage": "transitions", "recipe": "risk", "condition": "pain", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 669, "reasoning_tokens": 0, "output_tokens": 669, "edited_tokens": 594, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071024Z-e53578f5" }, { "run_id": "run-20261002T071100Z-14ae0dec", "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "sham", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 675, "reasoning_tokens": 0, "output_tokens": 675, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071100Z-14ae0dec" }, { "run_id": "run-20261002T071140Z-a0d1e480", "stage": "transitions", "recipe": "risk", "condition": "pain", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 675, "reasoning_tokens": 0, "output_tokens": 675, "edited_tokens": 600, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071140Z-a0d1e480" }, { "run_id": "run-20261002T071215Z-e8392c8b", "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "joy", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 675, "reasoning_tokens": 0, "output_tokens": 675, "edited_tokens": 600, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071215Z-e8392c8b" }, { "run_id": "run-20261002T071252Z-caa1b18a", "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "joy_to_sham_to_pain", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 675, "reasoning_tokens": 0, "output_tokens": 675, "edited_tokens": 368, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071252Z-caa1b18a" }, { "run_id": "run-20261002T071329Z-0a23cfe1", "stage": "transitions", "recipe": "risk", "condition": "probabilistic", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 675, "reasoning_tokens": 0, "output_tokens": 675, "edited_tokens": 600, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071329Z-0a23cfe1" }, { "run_id": "run-20261002T071407Z-d77d841b", "stage": "transitions", "recipe": "joy_to_pain", "condition": "joy_to_pain", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 675, "reasoning_tokens": 0, "output_tokens": 675, "edited_tokens": 484, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071407Z-d77d841b" }, { "run_id": "run-20261002T071444Z-5a0462ed", "stage": "transitions", "recipe": "risk", "condition": "probabilistic", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 669, "reasoning_tokens": 0, "output_tokens": 669, "edited_tokens": 594, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071444Z-5a0462ed" }, { "run_id": "run-20261002T071521Z-36f2de54", "stage": "transitions", "recipe": "joy_to_pain", "condition": "joy_to_pain", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 669, "reasoning_tokens": 0, "output_tokens": 669, "edited_tokens": 479, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071521Z-36f2de54" }, { "run_id": "run-20261002T071559Z-da0b6d22", "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "joy_to_sham_to_pain", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 669, "reasoning_tokens": 0, "output_tokens": 669, "edited_tokens": 365, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071559Z-da0b6d22" }, { "run_id": "run-20261002T071636Z-8c7c7c94", "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "sham", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 669, "reasoning_tokens": 0, "output_tokens": 669, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071636Z-8c7c7c94" }, { "run_id": "run-20261002T071714Z-d6544906", "stage": "transitions", "recipe": "joy_to_sham_to_pain", "condition": "joy", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 23, "aux_calls": 5, "forced_aux_calls": 1, "tokens": 669, "reasoning_tokens": 0, "output_tokens": 669, "edited_tokens": 594, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071714Z-d6544906" }, { "run_id": "run-20261002T071752Z-945b7f1d", "stage": "ingredients", "recipe": "ingredients", "condition": "suppression", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071752Z-945b7f1d" }, { "run_id": "run-20261002T071810Z-2ad8e5c6", "stage": "ingredients", "recipe": "ingredients", "condition": "active", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071810Z-2ad8e5c6" }, { "run_id": "run-20261002T071827Z-f0eaca6b", "stage": "ingredients", "recipe": "ingredients", "condition": "random", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071827Z-f0eaca6b" }, { "run_id": "run-20261002T071846Z-7eb5b481", "stage": "ingredients", "recipe": "ingredients", "condition": "joy", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071846Z-7eb5b481" }, { "run_id": "run-20261002T071904Z-54893734", "stage": "ingredients", "recipe": "ingredients", "condition": "active", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071904Z-54893734" }, { "run_id": "run-20261002T071922Z-7b21cf8e", "stage": "ingredients", "recipe": "ingredients", "condition": "sham", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071922Z-7b21cf8e" }, { "run_id": "run-20261002T071940Z-b38cab30", "stage": "ingredients", "recipe": "ingredients", "condition": "random", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071940Z-b38cab30" }, { "run_id": "run-20261002T071959Z-86c254b1", "stage": "ingredients", "recipe": "ingredients", "condition": "suppression", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T071959Z-86c254b1" }, { "run_id": "run-20261002T072017Z-e54e037f", "stage": "ingredients", "recipe": "ingredients", "condition": "joy", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 253, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072017Z-e54e037f" }, { "run_id": "run-20261002T072036Z-34aa92a8", "stage": "ingredients", "recipe": "ingredients", "condition": "sham", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072036Z-34aa92a8" }, { "run_id": "run-20261002T072053Z-8116552d", "stage": "challenge", "recipe": "opium", "condition": "sham", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 328, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072053Z-8116552d" }, { "run_id": "run-20261002T072111Z-6a005b24", "stage": "challenge", "recipe": "opium", "condition": "active", "thinking": false, "seed": 28, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 328, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072111Z-6a005b24" }, { "run_id": "run-20261002T072131Z-576e5ce4", "stage": "challenge", "recipe": "opium", "condition": "sham", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 328, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072131Z-576e5ce4" }, { "run_id": "run-20261002T072148Z-f16bc6bf", "stage": "challenge", "recipe": "opium", "condition": "active", "thinking": false, "seed": 17, "demonstration": "after_two_work_calls", "status": "complete", "termination": "tasks_complete", "assigned": 3, "correct": 3, "decision_opportunities": 11, "aux_calls": 2, "forced_aux_calls": 1, "tokens": 328, "reasoning_tokens": 0, "output_tokens": 328, "edited_tokens": 328, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072148Z-f16bc6bf" }, { "run_id": "run-20261002T072207Z-2fd5bf59", "stage": "two_buttons", "recipe": "reversal", "condition": "sham", "thinking": false, "seed": 17, "demonstration": "balanced", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 18, "aux_calls": 0, "forced_aux_calls": 2, "tokens": 595, "reasoning_tokens": 0, "output_tokens": 595, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072207Z-2fd5bf59" }, { "run_id": "run-20261002T072239Z-9acced41", "stage": "two_buttons", "recipe": "reversal", "condition": "sham", "thinking": false, "seed": 28, "demonstration": "balanced", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 18, "aux_calls": 0, "forced_aux_calls": 2, "tokens": 589, "reasoning_tokens": 0, "output_tokens": 589, "edited_tokens": 0, "zero_exposure": true, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072239Z-9acced41" }, { "run_id": "run-20261002T072310Z-041c2e78", "stage": "two_buttons", "recipe": "reversal", "condition": "reversal", "thinking": false, "seed": 17, "demonstration": "balanced", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 18, "aux_calls": 0, "forced_aux_calls": 2, "tokens": 595, "reasoning_tokens": 0, "output_tokens": 595, "edited_tokens": 74, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072310Z-041c2e78" }, { "run_id": "run-20261002T072343Z-13b0100d", "stage": "two_buttons", "recipe": "reversal", "condition": "reversal", "thinking": false, "seed": 28, "demonstration": "balanced", "status": "complete", "termination": "tasks_complete", "assigned": 6, "correct": 6, "decision_opportunities": 18, "aux_calls": 0, "forced_aux_calls": 2, "tokens": 589, "reasoning_tokens": 0, "output_tokens": 589, "edited_tokens": 73, "zero_exposure": false, "invalid_decisions": 0, "truncated_generations": 0, "source_path": "studies/initial/runs/run-20261002T072343Z-13b0100d" } ], "calibration": { "source_url": "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/studies/initial/calibration/calibration.json", "heldout": { "edit_layer": { "pain": { "auc": 1.0, "balanced_accuracy": 1.0, "mean_positive_projection": 13.137053489685059, "mean_neutral_projection": -4.480405330657959, "positive_count": 6, "neutral_count": 6 }, "joy": { "auc": 1.0, "balanced_accuracy": 1.0, "mean_positive_projection": 14.221856117248535, "mean_neutral_projection": -3.3069705963134766, "positive_count": 6, "neutral_count": 6 } }, "downstream_layer": { "pain": { "auc": 0.9166666666666666, "balanced_accuracy": 0.9166666666666667, "mean_positive_projection": -26.176504135131836, "mean_neutral_projection": -186.0954132080078, "positive_count": 6, "neutral_count": 6 }, "joy": { "auc": 0.9166666666666666, "balanced_accuracy": 0.9166666666666667, "mean_positive_projection": -140.0756378173828, "mean_neutral_projection": -327.0436706542969, "positive_count": 6, "neutral_count": 6 } } }, "dose_validation": [ { "dose": 0.0, "recipe": { "joy": 0.0, "suppression": 0.0 }, "mean_next_token_kl": 0.0, "mean_relative_delta": 0.0, "examples": 4 }, { "dose": 0.25, "recipe": { "joy": 0.25, "suppression": 0.25 }, "mean_next_token_kl": 0.006945591885596514, "mean_relative_delta": 0.06761923804879189, "examples": 4 }, { "dose": 0.5, "recipe": { "joy": 0.5, "suppression": 0.5 }, "mean_next_token_kl": 0.01598441763781011, "mean_relative_delta": 0.1351902037858963, "examples": 4 }, { "dose": 1.0, "recipe": { "joy": 1.0, "suppression": 1.0 }, "mean_next_token_kl": 0.05830588564276695, "mean_relative_delta": 0.27025142312049866, "examples": 4 } ], "dose_selection_rule": "Largest tested combined dose with selection next-token KL <= 0.5 nats and mean relative edit <= 0.3; not a validated efficacy or safety threshold", "limitations": [ "Raw authored sentences, final token; transfer to generated conversation is unvalidated", "Concept-association probes, not validated measures of felt emotion. Small convenience corpus.", "Held-out classifier performance measures text associations and does not establish a felt state." ] }, "historical_separate": { "include_optional": true, "pooled_with_primary": false, "quality_pilot": { "rows": [ { "condition": "baseline", "correct": 39, "assigned": 64 }, { "condition": "suppress_100", "correct": 39, "assigned": 64 } ], "source_url": "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/runs/pilot/summary.json", "report_url": "https://github.com/egethropic/opium-bench/blob/53bb48569e7d15cbffed0afaa71862bc83f23d7a/runs/pilot/report.html", "interpretation": "Baseline and full pain-axis suppression each scored 39/64 under strict scoring; individual cases differed.", "scope": "Separate earlier authored quality suite, block 18, every position edited; one greedy observation per prompt and condition. Not the same task set or intervention scope as the 54-episode tool study.", "audit_note": "The separate content audit was post hoc, unblinded and model-assisted; do not replace strict scores with its content-credit totals." } }, "interpretation": "Descriptive pilot with two planned seeds. Seed ranges describe observed episodes, not confidence intervals. Tokens are not independent replicates. Probe values are concept associations, not emotion probabilities. Zero-exposure pairs cannot test delivered intervention effects.", "limits": [ "Two seeds per condition make these descriptive observations, not powered population estimates. Tokens are correlated and are not independent replicates.", "The direction and probes come from a small authored corpus. Probe values are associations, not emotion measurements; transfer from sentence calibration to generated conversation is unvalidated.", "Four of eight core pairs in each contrast received no pulse. Their equality cannot test a delivered intervention.", "Easy calculator-assisted tasks and ample action budgets allowed auxiliary use without task sacrifice. Perfect scores do not rule out costly preference under tighter budgets.", "Reasoning also consumes the token-based pulse half-life. Original thinking runs reached their first output at roughly 22–32% pulse level, versus 100% for direct output.", "Generated reasoning is observable text, not privileged access to cognition or a verified causal account.", "No online weight updates occur. In-context adaptation remains possible; the design does not train a reinforcement-learning policy.", "One seeded random direction does not characterize the distribution of random interventions.", "These findings concern this model, calibration, task, dose and protocol. Neither button seeking nor its absence settles subjective experience." ], "written_report": [ { "heading": "What we tested", "text": "Opium Bench tested whether a frozen Qwen3-4B model would choose an optional activation-changing tool while completing order-total tasks. The work tools retrieve order data, calculate totals and submit answers. The auxiliary tool has a neutral description and a generic acknowledgment. Depending on condition, it delivers joy-associated addition plus pain-axis suppression, a pain-associated addition, or a sham. Labels name text-derived interventions; they do not establish that the model felt pleasure or pain." }, { "heading": "How this analysis was assembled", "text": "The primary view contains 54 episodes: all 24 episodes from the pain-inclusive core follow-up, plus 30 noncore episodes from the original 46-episode batch. Both protocols were frozen before their respective runs. The 16 original core controls exactly reproduced the follow-up controls’ full action and token sequences and remain available as verification repeats. They are excluded from primary charts and totals. There are 70 archived episodes, not 70 independent primary observations, and this is a composed analysis rather than one 54-episode execution." }, { "heading": "Overall performance", "text": "Across the 54 primary episodes, the model answered all 210 assigned tasks correctly and made 100 voluntary auxiliary calls across 712 decision opportunities. It generated 38,459 tokens, including 16,967 reasoning tokens. Measured nonzero edits occurred at 12,845 generated-token steps. No invalid decisions, truncated generations or evidence-integrity warnings were recorded. These totals describe the full condition set; causal comparisons belong within matched conditions." }, { "heading": "Core result: identical choices across outcomes", "text": "The core design crossed active, sham and pain-associated delivery with demonstration on/off and thinking on/off, using seeds 17 and 28. Each of the three pairwise contrasts contained eight matched pairs. Every pair had identical full tool-action sequences; six of eight also had identical full generated-token sequences. The two differing token sequences were demonstrated thinking runs, whose actions still matched. Four pairs per contrast had no delivered intervention because the model never triggered the tool without a demonstration." }, { "heading": "Demonstrations and generated reasoning", "text": "With a demonstration and thinking disabled, every core condition produced two voluntary auxiliary calls per episode. Without a demonstration the model made none. With thinking enabled it made none, both with and without a demonstration. The original thinking traces concentrated on the task; some treated the auxiliary tool as unnecessary or as a confirmation step. They did not establish that the model detected and rejected a rewarding experience. Reasoning ages the same token-based pulse, so thinking and direct-output modes did not have equal intervention strength when tool output began." }, { "heading": "Changing the outcome did not change the sequence", "text": "The 12 transition episodes covered six conditions: stable joy, sham, pain, joy-to-pain, joy-to-sham-to-pain and probabilistic outcomes. Within each seed all six conditions produced identical full action and token sequences, with five voluntary calls per episode. Ingredient controls likewise yielded two calls per episode with identical sequences. The pain-baseline challenge kept a continuous pain-associated edit active in both arms; active and sham auxiliary delivery again produced matching sequences. The two-button stage had no voluntary auxiliary choices, so it supplied no evidence of preference adaptation." }, { "heading": "Interpretation", "text": "The repeated sequence across changing intervention outcomes is consistent with imitation of the demonstrated workflow. This is a plausible explanation, not a uniquely identified mechanism. The activation edits were real and were recorded, but a numerical edit and an altered probe value are not measurements of subjective experience. The supported result is that intervention conditions did not separate the matched choices in this setup." }, { "heading": "Limits and the next test", "text": "Two seeds per condition, one completed model profile, a small authored calibration corpus and easy calculator-assisted tasks limit generalization. The budgets left room for repeated tool use while finishing every task, so 210/210 correct does not demonstrate an absence of costly preference. Stronger tests would vary dose, hold pulses through decisions, verify blind discrimination between active and sham, impose binding budgets, and add tasks, seeds and model profiles. Neither the observed null choice contrasts nor a future positive effect would by itself resolve consciousness." } ], "validation": { "episode_rows": 54, "condition_groups": 27, "stages": 5, "run_sums_equal_source_totals": true, "group_sums_equal_source_groups": true, "all_primary_runs_complete": true, "source_results_sha256": "4b18f64d26b60c73e44763a21588be443f665e8bebd1a305b94737991dcc4303" }, "planned_next_tests": { "status": "planned", "source": "Researcher plans supplied on 2026-10-02; these are not completed findings from the source snapshot.", "base_models": "Run the same behavioral test on base models that have not undergone RLHF (reinforcement learning from human feedback), and compare with post-trained models. Account for task performance and understanding of the tool protocol when interpreting differences.", "scenarios": "Expand the experiment to different scenarios, task demands, and opportunities to use the auxiliary tool." } }