Today for AI
HOT RADAR
llmHEAT 12.7°

AI CURATED HOT STORY · 11/15/2026

Weak AI Models Can Grade Steps to Train Stronger Agents, Zhejiang University Finds

1 reports archived1 independent sourcesupdated 11/15/2026, 02:59:28
Synthesis & Latest Updates
1 Sources Cross-Validated

A new Zhejiang University paper shows that a decision model, JEV, which completes only 42.9% of ALFWorld tasks on its own, can still act as a step-level grader and lift another agent's success rate from 72.8% to 82.9%. The grader reviews a full trajectory after the attempt and its per-action scores are folded into GRPO training, sidestepping the uniform credit assignment of outcome-only rewards and the cost of per-environment labels or per-action LLM judges. The authors argue step graders should be selected by how well they judge recorded runs, not by how well they act.

LATEST/A new Zhejiang University paper shows that a decision model, JEV, which completes only 42.9% of ALFWorld tasks on its own, can still act as a step-level grader and lift another agent's success rate from 72.8% to 82.9%. The grader reviews a full trajectory after the attempt and its per-action scores are folded into GRPO training, sidestepping the uniform credit assignment of outcome-only rewards and the cost of per-environment labels or per-action LLM judges. The authors argue step graders should be selected by how well they judge recorded runs, not by how well they act.

Hourly heat curve

2 fully observed hours
Fewer than 3 fully observed hours; trend not plotted yet.

Coverage timeline

Total 1 reports · Latest first
  1. Rohan Paul (@rohanpaul_ai)T2·71 pts

    Weak AI Models Can Grade Steps to Train Stronger Agents, Zhejiang University Finds

    • JEV solves only 42.9% of ALFWorld tasks alone but as a step grader raises a Qwen2.5-1.5B policy from 72.8% to 82.9% success
    • Per-step scores are assigned after the full trajectory and folded directly into GRPO, avoiding per-action LLM judges or environment labels
    • Graders should be chosen for how well they judge recorded runs, not how well they act