Recent systems train coding agents with reinforcement learning (RL) on trajectories collected through multiple execution harnesses. However, gains under the training harnesses leave open whether the improvements transfer to unseen interfaces and how they depend on reward normalization. We present a controlled study tha...
Chen-Qian Le, Jia-Yi Cheng, Qi-Jia He et al.· 1 citation
This work forms a post-failure decision as recovery routing over heterogeneous actions and train a supervised router from execution rollouts to make the same router usable under changing budgets, and adds a Conformal Risk Control layer that selects a deployment-time cost penalty without retraining and provides marginal...
Qi-Jia He, Jia-Yi Cheng, Chen-Qian Le et al.· arXiv.org· 2 citations
Agent reinforcement learning (RL) increasingly runs through full execution harnesses, and a multi-harness recipe mixes two choices: exposing the policy to several harnesses, and comparing their rewards inside one relative-advantage group. We isolate the second choice in repository-level coding. From one Qwen3-8B superv...
Chen-Qian Le, Jia-Yi Cheng, Qi-Jia He et al.· 1 citation
This method generates candidate responses from the target policy, evaluates helpfulness, factuality, and conciseness with rubric-specialized evaluators, applies a process-critic correction, and retains only high-consensus desirable or undesirable examples.
Bounded unfiltered teacher continuations at learner-induced contexts improve over pure behavioral cloning at matched budgets and suggest that a few teacher steps, placed at learner-induced contexts, can be a more cost-efficient supervision allocation than longer or more heavily curated teacher completions.
Junze Ye, Jiayi Cheng, Miao Lu et al.· arXiv.org· 2 citations
PreDiff-LM preserves causal attention within the observed prompt while allowing full bidirectional attention within the masked target, position hybrid attention as a complementary mechanism for adapting pretrained causal backbones, while making explicit the remaining quality and inference-efficiency gaps to optimized A...
