Cell · cited by the thesis

simple_qwen3.6-35b_final2_la3

Qwen3.6-35B on the simple floor, 10 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)7/10 [0.40, 0.89]
Final HP / exact optimum, mean over wins15%
Died0
Final HP, mean161
Seconds per model call20.4
Parse error rate0.0%
Settings
FloorSimple floor
ModelQwen3.6-35B (qwen3.6-35b-a3b, thinking)
Planningreactive
Promptengine-default-floor-basic
Temperature0
Lookahead depth3
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
1won215%3617replay
2won215%3617replay
3won215%3617replay
4not won526–3617replay
5won215%3617replay
6won215%3617replay
7won215%3617replay
8not won545–3617replay
9won215%3617replay
10not won526–3617replay
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_qwen3.6-35b_final2_la3, never over the thesis result):

python scripts/run_catalog.py --codes G1s

Result file: results/reliability/simple_qwen3.6-35b_final2_la3.json · updated 2026-09-27