Cell · cited by the thesis

simple_qwen3.6-35b_final2_la0

Qwen3.6-35B on the simple floor, 5 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)0/5 [0.00, 0.43]
Final HP / exact optimum, mean over wins–
Died0
Final HP, mean485
Seconds per model call22.5
Parse error rate0.0%
Settings
FloorSimple floor
ModelQwen3.6-35B (qwen3.6-35b-a3b, thinking)
Planningreactive
Promptengine-default-floor-basic
Temperature0
Lookahead depth0
Mechanical overrideson (logged per turn)
Engine settings (environment, 4)
THINKING_MAX_TOKENS=16384
CONVERSATION_HISTORY_TURNS=0
MOTA_ORDER_TWO_SIDED=1
MOTA_FIGHT_KEY_FLOOD=1
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won470–3616replay
2not won477–3616replay
3not won473–3615replay
4not won470–3616replay
5not won533–3616replay
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_qwen3.6-35b_final2_la0, never over the thesis result):

python scripts/run_catalog.py --codes G3bs

Result file: results/reliability/simple_qwen3.6-35b_final2_la0.json · updated 2026-09-27