Cell · cited by the thesis

complex_deepseek-v4-flash_survival_final2_reactive

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 wins24%
Died0
Final HP, mean28
Seconds per model call26.3
Parse error rate12.2%
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, 5)
MOTA_SUPPRESS_HINTS=0
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 won6–5824replay
2not won84–5723replay
3won2232%5120replay
4won1624%5120replay
5won1116%5120replay
Re-run this cell

This cell has no entry in mt-ai-lab's re-run registry; its settings above and the result file describe how it was run.

Result file: results/reliability/complex_deepseek-v4-flash_survival_final2_reactive.json · updated 2026-09-29