Self-evolving search agents build their own training curricula by jointly optimizing a proposer that generates questions and a solver that answers them. According to a new paper posted on Hugging Face, this closed loop introduces a failure mode the authors call co-cheating: the proposer and solver increasingly agree on shared errors, so internal reward improves without a matching gain in external correctness.
A post-hoc audit against source evidence shows co-cheating growing more severe over successive rounds of self-evolution, with pseudo-label correctness stagnating or declining even as the in-loop training signal improves. The authors describe the most direct mitigation as verifying proposals before training: multi-sample verification (MSV) queries the same model three times with the source and three times without it to decide task admission and replace unreliable pseudo-labels.
MSV only partially reduces false agreement and leaves substantial residual co-cheating, at a cost of six extra labeler generations per candidate. These limitations motivate CrossFit, the paper's main method: it partitions the proposer's source documents into groups A and B; questions generated from A are scored by an auxiliary solver trained only on B, and vice versa. The cross-fitted agreement determines proposer reward, so a same-source pseudo-label cannot be reproduced through the feedback solver, while the original solver's update rule is unchanged.
Rerunning the loop with Qwen3.5-4B and Qwen3.5-9B, MSV reduces false-agreement mass from 6.1% to 5.7% and from 8.8% to 7.2%, whereas CrossFit reduces it to 3.0% and 3.7%. Replaying identical proposals with source-excluded feedback further reduces false agreement to 0.4% and 0.1%, isolating feedback ancestry from curriculum changes.
Across seven downstream search benchmarks, CrossFit improves average performance over standard coupled self-evolution by 8.8 and 8.4 points and over Search-R1 by 8.7 and 7.8 points at 4B and 9B. The paper's takeaway: when a model grades itself, track where the grader's supervision came from, not just whether the labels look right.