Cell · cited by the thesis

complex_deepseek-v4-flash_survival_final2_reactive_t32_t10

DeepSeek-V4-Flash on the complex floor, 5 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)2/5 [0.12, 0.77]
Final HP / exact optimum, mean over wins24%
Died0
Final HP, mean26
Seconds per model call16.2
Parse error rate0.0%
Settings
FloorComplex floor
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptcomplex_bfs_strategy_survival.txt
Temperature1
Lookahead depth3
Mechanical overridesoff (all engine replacements and assists disabled)
Engine settings (environment, 6)
MOTA_SUPPRESS_HINTS=0
THINKING_MAX_TOKENS=32768
CONVERSATION_HISTORY_TURNS=0
MOTA_ORDER_TWO_SIDED=1
MOTA_FIGHT_KEY_FLOOD=1
LLM_TIMEOUT=1800
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won3–5624replay
2won1624%5120replay
3not won79–5723replay
4not won15–5723replay
5won1624%5120replay
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>_complex_deepseek-v4-flash_survival_final2_reactive_t32_t10, never over the thesis result):

python scripts/run_catalog.py --codes T10DCR

Result file: results/reliability/complex_deepseek-v4-flash_survival_final2_reactive_t32_t10.json · updated 2026-09-30