1experiment_id: E7_qtypes2splits:3- test4families:5- synthetic6lengths:7- 20008- 80009- 1600010limit_docs_per_family: 4011question_range:12- 513- 1514model_pairs:15- sender: qwen3b16 receiver: qwen3b17- sender: llama3b18 receiver: llama3b19generation:20 max_tokens: 4821seeds:22- 023protocol_hash: 404c52b34d964693da72ef459a036ffc42fbf5f6ba8748cc9de0c554ebebddd624note: 'EXPLORATORY (amendment A-004): question types 6-15 (negation, revision, temporal,25 aggregation, unit conversion, unanswerable) not covered by the 5-question core limit'26conditions:27- method: A128- method: A429 budget_tokens: 51230 access: blind31- method: A532 budget_tokens: 51233 access: blind34- method: A335 budget_tokens: 51236 access: blind37- method: B338 keep_frac: 0.2539- method: B540 keep_frac: 0.2541 access: aware42- method: B243 bits: 844