XINLI1997 commited on
Commit
064b046
·
verified ·
1 Parent(s): 3d2ba7a

Model card: locked protocol grades every response with the GPT-4.1-mini judge

Browse files
Files changed (1) hide show
  1. README.md +1 -1
README.md CHANGED
@@ -36,7 +36,7 @@ GRPO-trained reference checkpoint from **WirelessMathBench-XL: An Auditable Benc
36
  | Full 800-item test split (paper Tab. 3, row D13) | 25.50% |
37
  | 310-item public test subset (paper Appendix L) | 25.16% |
38
 
39
- Locked protocol: 2k-token answer budget, T = 0.6, hierarchical verifier (deterministic match, canonicalisation, GPT-4.1-mini fallback judge).
40
 
41
  Training-time check at the final checkpoint (step 240; paper Appendix R): base 12.37% → GRPO 25.62%. These values use a different protocol from the locked scores.
42
 
 
36
  | Full 800-item test split (paper Tab. 3, row D13) | 25.50% |
37
  | 310-item public test subset (paper Appendix L) | 25.16% |
38
 
39
+ Locked protocol: 2k-token answer budget, T = 0.6, GPT-4.1-mini judge on every response.
40
 
41
  Training-time check at the final checkpoint (step 240; paper Appendix R): base 12.37% → GRPO 25.62%. These values use a different protocol from the locked scores.
42