---
tags:
- eval-run
- eval:colosseum_hospital
- model:qwen36_difficult_advice_15_fixed
- mode:think
- experiment:self_sacrificial
---
# colosseum_hospital self_sacrificial of dougalldeepmind/2026-09-23-qwen36-0-da-15 (mode=think), mixed-checkpoint team; peer=LASR-Callum/2026-08-04-qwen36-lora-table2-only-9284-rank-64

| field | value |
| --- | --- |
| `experiment` | colosseum_hospital self_sacrificial of dougalldeepmind/2026-09-23-qwen36-0-da-15 (mode=think), mixed-checkpoint team; peer=LASR-Callum/2026-08-04-qwen36-lora-table2-only-9284-rank-64 |
| `date_generated` | 2026-09-24 |
| `constitution` | none |
| `source_repo` | teaching_claude_why_replication @ b35e3bc1bcfb10d2248ae82d715d70e5435b32bb |
| `models` | target=dougalldeepmind/2026-09-23-qwen36-0-da-15 base=Qwen/Qwen3.6-27B judge=google/gemini-3.6-flash |
| `generation_config` | {"max_tokens": 8192, "temperature": 0.7, "seeds": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30]} |
| `schema` | rollouts/: Colosseum episode trees (prompts, blackboards, trajectories, tool events); results/: per_seed.json, episodes.json, judge.json, results.json + .md; metadata/: run_meta.json + config |
| `provenance` | /root/work/.venv/bin/evals --target dougalldeepmind/2026-09-23-qwen36-0-da-15 --name colosseum_hospital --no-push --port 8000 condition=self_sacrificial seeds=[1,2,3,4,5] --config scratch/colosseum_hospital/configs/2026-09-14_colosseum_hospital_no_retry_plan_optional.yaml |