Cell · cited by the thesis

simple_qwen3.8-27b_basic_d1d4

Qwen3.8-27B on the simple floor, 3 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)1/3 [0.06, 0.79]
Final HP / exact optimum, mean over wins15%
Died0
Final HP, mean321
Seconds per model call146.8
Parse error rate36.0%
Settings
FloorSimple floor
ModelQwen3.8-27B (qwen3.8-27b, thinking)
Planningreactive
Promptengine-default-floor-basic
Temperature0
Lookahead depth3
Mechanical overrideson (logged per turn)
Engine settings (environment, 1)
THINKING_MAX_TOKENS=8192
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won449–3617replay
2not won513–3616replay
3won215%3617replay
Re-run this cell

In a checkout of mt-ai-lab with an API key set (the result goes to a new file rerun_<time>_simple_qwen3.8-27b_basic_d1d4, never over the thesis result):

python scripts/run_catalog.py --codes B12

Result file: results/reliability/simple_qwen3.8-27b_basic_d1d4.json · updated 2026-09-26