Cell · cited by the thesis

complex_evolved_route_deepseek-v4-flash

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

Outcome
Wins (boss defeated, alive)0/5 [0.00, 0.43]
Final HP / exact optimum, mean over wins–
Died0
Final HP, mean76
Seconds per model call27.2
Parse error rate7.6%
Settings
FloorComplex floor
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptevolved_route_deepseek.txt
Temperature0
Lookahead depth3
Mechanical overrideson (logged per turn)
Engine settings (environment, 5)
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 won93–6122replay
2not won23–5921replay
3not won107–3514replay
4not won155–3313replay
5not won2–229replay
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_evolved_route_deepseek-v4-flash, never over the thesis result):

python scripts/run_catalog.py --codes ERD

Result file: results/reliability/complex_evolved_route_deepseek-v4-flash.json · updated 2026-10-05