← older: smoke__qwen3.6-35b-a3b__20260726-002616all runs · task boardnewer: smoke__qwen3.6-35b-a3b__20260726-034536 →

smokeqwen3.6-35b-a3b20260726-025257

No notes for this run yet. Add runs/smoke__qwen3.6-35b-a3b__20260726-025257/NOTES.md — what was run & what changed, results, interpretation, next steps — and it will render here.

Run details

modelllama-local/qwen3.6-35b-a3bagentharnesses.minimal_pi:MinimalPithinkingonreasoning budgetdirect (server/none)* — see journal for the authoritative mechanismmaxTokens / contextWindow65536 / 196608agent timeout ×2.0trials9 of 9 — 5 pass · 4 failmean reward0.56tokens (job total)18,564,633 in / 178,267 outstarted / finished2026-07-26T02:52 / 2026-07-26T03:45wall clock52m36s

Tasks

cobol-modernization — 3/3 passed

#	result	total	agent	in/out tok	flags
1	PASS	8m13s	7m39s	7809309/70588	long reasoning (25,677 chars)loop-guard blocked a repeated call ×2	🔍 view
2	PASS	3m30s	2m57s	808357/29961	long reasoning (22,304 chars)	🔍 view
3	PASS	1m39s	1m06s	471844/10804	—	🔍 view

crack-7z-hash — 2/3 passed

#	result	total	agent	in/out tok	flags
1	PASS	3m18s	2m38s	246682/3118	—	🔍 view
2	PASS	2m46s	2m07s	203585/3074	—	🔍 view
3	FAIL	24m06s	20m00s	806987/11602	a bash command timed out ×4fast-timeout cut at 20m	🔍 view

sanitize-git-repo — 0/3 passed

#	result	total	agent	in/out tok	flags
1	FAIL	3m29s	2m55s	3162399/16736	claimed success but the verifier did NOT pass (heuristic)	🔍 view
2	FAIL	1m47s	1m13s	1068607/8286	—	🔍 view
3	FAIL	3m45s	3m11s	3986863/24098	claimed success but the verifier did NOT pass (heuristic)	🔍 view