7B students on mathematical reasoning and code generation.
Difficulty-gated teacher guidance improves solution coverage in on-policy training
DiffGate supplements reward-based training with bounded token-level teacher signals only on failed trajectories, producing consistent pass@8 gains over matched GRPO.
Industry
Karn Tiwari · Varnith Chordia · Prathosh A P
Indian Institute of Science · Snap Inc.
Research Digest··2 min read
P.
Why this paper
From Indian Institute of Science and Snap Inc.
In one line
DiffGate applies bounded teacher guidance to failed trajectories only, improving pass@8 over GRPO across Qwen3 students in code and math.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§