Cell · cited by the thesis

complex_deepseek-v4-flash_survival_final2_plan3

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

Outcome
Wins (boss defeated, alive)3/5 [0.23, 0.88]
Final HP / exact optimum, mean over wins26%
Died0
Final HP, mean25
Seconds per model call88.5
Parse error rate4.2%
Settings
FloorComplex floor
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptcomplex_bfs_strategy_survival.txt
Temperature0
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
MOTA_PLAN_STEPS=3
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won49–5712replay
2not won20–5512replay
3won1015%518replay
4won2232%528replay
5won2232%518replay
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_plan3, never over the thesis result):

python scripts/run_catalog.py --codes DCP3 DCP3a

Result file: results/reliability/complex_deepseek-v4-flash_survival_final2_plan3.json · updated 2026-09-29