Search papers, labs, and topics across Lattice.
ReflexiCoder, a novel reinforcement learning framework, trains LLMs to internalize a structured reasoning trajectory of code generation, bug-aware reflection, and self-correction. The framework uses an RL-zero training paradigm with granular reward functions to optimize the entire reflection-correction trajectory without relying on ground-truth feedback or execution engines at inference time. ReflexiCoder-8B achieves state-of-the-art results among open-source models in the 1.5B-14B range on HumanEval, MBPP, BigCodeBench, LiveCodeBench, and CodeForces, while also reducing inference-time compute overhead by approximately 40%.
Forget external debuggers: ReflexiCoder teaches LLMs to self-reflect and self-correct code, rivaling GPT-5.1 in performance while slashing inference costs by 40%.
While Large Language Models (LLMs) have revolutionized code generation, standard"System 1"approaches, generating solutions in a single forward pass, often hit a performance ceiling when faced with complex algorithmic tasks. Existing iterative refinement strategies attempt to bridge this gap at inference time, yet they predominantly rely on external oracles, execution feedback, or computationally expensive prompt-response cycles. In this work, we propose ReflexiCoder, a novel reinforcement learning (RL) framework that internalizes the structured reasoning trajectory, encompassing initial generation, bug and optimization aware reflection, and self-correction, directly into the model's weights. Unlike prior methods, ReflexiCoder shifts the paradigm from external-dependent refinement to an intrinsic, fully autonomous self-reflection and self-correction capabilities at inference time. We utilize an RL-zero training paradigm with granular reward functions to optimize the entire reflection-correction trajectory, teaching the model how to debug without reliance on ground-truth feedback or execution engines at inference time. Extensive experiments across seven benchmarks demonstrate that our ReflexiCoder-8B establishes a new state-of-the-art (SOTA) among leading open-source models in the 1.5B-14B range, achieving 94.51% (87.20%) on HumanEval (Plus), 81.80% (78.57%) on MBPP (Plus), 35.00% on BigCodeBench, 52.21% on LiveCodeBench, and 37.34% on CodeForces in a single-attempt setting, rivaling or surpassing proprietary models like GPT-5.1. Notably, our framework is significantly more token-efficient than base models, reducing inference-time compute overhead by approximately 40% through disciplined, high-speed reasoning and reflection patterns. Source code is available at https://github.com/juyongjiang/ReflexiCoder.
