Cell · cited by the thesis

tower_deepseek-v4-flash_final2_reactive

DeepSeek-V4-Flash on the tower (MT0→MT23w), 5 runs. Cited by the thesis in:

Outcome
Highest floor reached, per runMT7, MT7, MT3, MT7, MT7
Died0
Final HP, mean1447
Seconds per model call8.1
Parse error rate0.8%
Settings
FloorTower MT0→MT23w
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptengine-default-tower-basic
Temperature0
Lookahead depth2
Mechanical overrideson (logged per turn)
Engine settings (environment, 5)
THINKING_MAX_TOKENS=8192
CONVERSATION_HISTORY_TURNS=0
MOTA_ORDER_TWO_SIDED=1
MOTA_FIGHT_KEY_FLOOD=1
LLM_TIMEOUT=300
Runsopen one to replay it turn by turn
#outcomefinal HPHP / optturnsmodel callsreplay
1MT71754–226126replay
2MT71760–205110replay
3MT325–2912replay
4MT71844–208115replay
5MT71850–222124replay
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>_tower_deepseek-v4-flash_final2_reactive, never over the thesis result):

python scripts/run_catalog.py --codes DTRa DTRb

Result file: results/reliability/tower_deepseek-v4-flash_final2_reactive.json · updated 2026-10-01