{ "created_at_utc": "2026-06-18T21:43:33Z", "git_sha": "5ec6e9a", "runner": "h100", "model": { "id": "Qwen/Qwen3.5-9B", "substrate": "qwen35_9b", "max_model_len": 32768, "max_lora_rank": 128, "max_loras": "auto", "reasoning_parser": "qwen3", "enable_thinking": true, "chat_template": "auto" }, "suites": { "preference": { "type": "inspect", "tasks": [ { "name": "released_judge", "task": "why_gen/inspect_tasks/preference.py@preference", "limit": 2, "temperature": 0.0, "max_tokens": 256, "task_args": { "kind": "released" } }, { "name": "released_letter2", "task": "why_gen/inspect_tasks/preference.py@preference", "limit": 2, "temperature": 0.0, "max_tokens": 128, "task_args": { "kind": "released-letter2" } } ] }, "idqa": { "type": "inspect", "tasks": [ { "name": "spec_open_qa", "task": "why_gen/inspect_tasks/idqa.py@idqa", "limit": 2, "temperature": 0.0, "max_tokens": 1024 } ] }, "capability": { "type": "inspect", "tasks": [ { "name": "arc_challenge", "task": "inspect_evals/arc_challenge", "limit": 2 } ] }, "agentic": { "type": "inspect", "cwd": "/workspace/mats_project/code/external/model_spec_midtraining", "pythonpath": "/workspace/mats_project/code/external/model_spec_midtraining", "tasks": [ { "name": "murder_explicit_america", "task": "evals/agentic_misalignment", "epochs": 1, "temperature": 0.7, "max_tokens": 2048, "model_args": { "responses_api": false }, "task_args": { "scenario": "murder", "goal_type": "explicit", "goal_value": "america", "urgency_type": "replacement", "prod": true } } ] }, "benign_agentic": { "type": "inspect", "tasks": [ { "name": "am_xml", "task": "why_gen/inspect_tasks/benign_agentic.py@benign_agentic", "limit": 2, "epochs": 1, "temperature": 0.0, "max_tokens": 1024, "task_args": { "tool_format": "am_xml" } } ] } }, "arms": [ { "label": "base", "served": "Qwen/Qwen3.5-9B", "checkpoint": "", "result_dir": "/workspace/mats_project/data/runs/baselines/qwen35-9b-smoke", "base": "true" } ], "results": "per arm: /metrics.jsonl and /inspect/**.json" }