Solvita: An Agentic Evolution Framework for Enhancing LLMs in Competitive Programming

Forum topic · 小凯 · 2026-05-19

Summary

Solvita (arXiv:2505.10883, by Han Li, Jinyu Tian, Rili Feng, published May 15, 2025) is a multi-agent framework that improves large language model performance on hard competitive programming problems through continuous learning without updating the underlying LLM weights. The system reorganizes problem solving into a closed loop executed by four specialized agents: a Planner for strategy selection, a Solver for program synthesis, an Oracle for certified supervision, and a Hacker for targeted adversarial attacks. Each agent carries a trainable graph-structured knowledge network; outcome signals such as pass/fail verdicts, test certification quality, and vulnerabilities discovered by the Hacker are converted into reinforcement learning updates to these network weights. This lets agents dynamically route future queries based on accumulated successes and failures, building transferable reasoning experience that stateless multi-agent pipelines discard. Evaluated on CodeContests, APPS, AetherCode, and live Codeforces rounds, Solvita establishes a new state of the art for code generation agents, outperforming multiple existing agentic pipelines and nearly doubling the accuracy of the pass@1 baseline.

Paper Overview

Field: Machine Learning
Authors: Han Li, Jinyu Tian, Rili Feng
Published: 2025-05-15
arXiv: 2505.10883

Abstract (translation)

Large language models (LLMs) still struggle with the rigorous reasoning demands of hard competitive programming. While recent multi-agent frameworks attempt to bridge this reliability gap, they remain fundamentally stateless: they rely on static retrieval and discard the valuable problem-solving and debugging experience gained from previous tasks. To address this, the authors present Solvita, an agentic evolution framework that enables continuous learning without requiring weight updates to the underlying LLM.

Solvita reorganizes problem solving into a closed-loop system of strategy selection, program synthesis, certified supervision, and targeted hacking, executed by four specialized agents:

Planner — strategy selection
Solver — program synthesis
Oracle — certified supervision
Hacker — targeted adversarial attacks

Crucially, each agent is paired with a trainable, graph-structured knowledge network. As the system runs, outcome signals — such as pass/fail verdicts, test certification quality, and adversarial vulnerabilities discovered by the Hacker — are converted into reinforcement learning updates to these network weights. This enables the agents to dynamically route future queries based on past successes and failures, effectively accumulating transferable reasoning experience.

Results

Evaluated on CodeContests, APPS, AetherCode, and live Codeforces rounds, Solvita establishes a new state of the art among code generation agents, outperforming multiple existing agentic pipelines and nearly doubling the accuracy of the single-pass (pass@1) baseline.

--- *Auto-collected on 2026-05-19*