Paper 2606.28436
Dockerless: Environment-Free Program Verifier for Coding Agents
- Published
- Jun 2026
- Research lab
- Independent
- Citations
- 1
- GitHub
- Not linked
01 In brief
Summary
Dockerless is an environment-free agentic program verifier that evaluates code patches without executing them, using agentic repository exploration to gather evidence.
It generates verification questions from the issue and reference patch, dispatches parallel sub-agents to collect evidence-backed answers, and produces a correctness score.
Trained via rejection sampling on 3.7K issues, Dockerless outperforms the strongest open-source verifier by 14.3 AUC points.
It enables a fully environment-free post-training pipeline for SFT trajectory filtering and RL rewards.
The resulting model achieves 62.0%, 50.0%, and 35.2% resolve rates on SWE-bench Verified, Multilingual, and Pro, surpassing the Qwen3.5-9B baseline by 2.4, 8.7, and 2.9 points, matching environment-based post-training performance.
Env-free SFT and RL match env-based counterparts, with Dockerless filtering 4K top trajectories outperforming random selection.
The verifier's performance improves with up to 4 verification questions, and its latency overhead is minimal (7.2% of total RL time).
A case study shows Dockerless correctly scores a patch that differs in surface form from the reference, demonstrating its repository-grounded judgment.
02 From the paper
Abstract
Program verifiers play a central role in training coding agents, including selecting trajectories for supervised fine-tuning (SFT) and providing rewards for reinforcement learning (RL). Standard execution-based verification requires running unit tests inside per-repository environments such as Docker images, incurring substantial environment setup costs. We propose Dockerless, an environment-free agentic patch verifier that evaluates generated code patches without executing them. Rather than simply matching candidate patches to references, Dockerless judges patch correctness using evidence gathered through agentic repository exploration. On a verifier evaluation benchmark, Dockerless outperforms the strongest open-source verifier by 14.3 AUC points. Using Dockerless as both the SFT trajectory filter and the RL reward enables a fully environment-free post-training pipeline. The resulting model reaches 62.0%, 50.0%, and 35.2% resolve rate on SWE-bench Verified, Multilingual, and Pro, respectively. It surpasses the Qwen3.5-9B baseline by 2.4, 8.7, and 2.9 points, matching environment-based post-training.