Meta-Eval/EvalHarness
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

aime/hmmt max_tokens 32768→8192: stays on non-stream path where GLM gateway honors enable_thinking:false (thinking off, 15s/题)

Browse Source 
Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
sora 2026-09-11 09:43:36 +00:00
parent 3f555878b8
commit 624b2fd37e
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
evalharness/config/default.yaml
View File

@ -6,15 +6,19 @@ default:
aime24: aime24:
temperature: 1.0 temperature: 1.0
repeats: 12 repeats: 12
max_tokens: 8192
aime25: aime25:
temperature: 1.0 temperature: 1.0
repeats: 12 repeats: 12
max_tokens: 8192
aime26: aime26:
temperature: 1.0 temperature: 1.0
repeats: 12 repeats: 12
max_tokens: 8192
hmmt26: hmmt26:
temperature: 1.0 temperature: 1.0
repeats: 12 repeats: 12
max_tokens: 8192
imo_answerbench: imo_answerbench:
temperature: 1.0 temperature: 1.0
gpqa_diamond: gpqa_diamond: