Cell

simple_qwen3.5-397b_basic_k1k2_hist0_compact

Qwen3.5-397B on the simple floor, 10 episodes. Not cited by the thesis (development or earlier run).

Outcome
Wins (boss defeated, alive)9/10 [0.60, 0.98]
Final HP / exact optimum, mean over wins15%
Died0
Final HP, mean57
Seconds per model call46.4
Parse error rate0.0%
Settings
FloorSimple floor
ModelQwen3.5-397B (qwen3.5-397b-a17b, thinking)
Planningreactive
Promptengine-default-floor-basic
Temperature0
Lookahead depth3
Mechanical overrideson (logged per turn)
Engine settings (environment, 3)
THINKING_MAX_TOKENS=32768
CONVERSATION_HISTORY_TURNS=0
MOTA_COMPACT_PROMPT=1
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1won215%3617replay
2won215%3617replay
3won215%3617replay
4won215%3617replay
5won215%3617replay
6won215%3617replay
7won215%3617replay
8won215%3617replay
9not won550–3617replay
10won215%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_qwen3.5-397b_basic_k1k2_hist0_compact, never over the thesis result):

python scripts/run_catalog.py --codes CP1

Result file: results/reliability/simple_qwen3.5-397b_basic_k1k2_hist0_compact.json · updated 2026-09-27