6 papers
LK Losses: Direct Acceptance Rate Optimization for Speculative Decoding
Alexander Samarin, Sergei Krutikov, Anton Shevtsov +3
Speculative decoding accelerates autoregressive large language model (LLM) inference by using a lightweight draft model to propose candidate tokens that are then verified in parall…
SWE-rebench V2: Language-Agnostic SWE Task Collection at Scale
Ibragim Badertdinov, Maksim Nekrashevich, Anton Shevtsov +1
Software engineering agents (SWE) are improving rapidly, with recent gains largely driven by reinforcement learning (RL). However, RL training is constrained by the scarcity of lar…
Blockwise Advantage Estimation for Multi-Objective RL with Verifiable Rewards
Kirill Pavlenko, Alexander Golubev, Simon Karasik +1
Group Relative Policy Optimization (GRPO) assigns a single scalar advantage to all tokens in a completion. For structured generations with explicit segments and objectives, this co…
SWE-rebench: An Automated Pipeline for Task Collection and Decontaminated Evaluation of Software Engineering Agents
Ibragim Badertdinov, Alexander Golubev, Maksim Nekrashevich +6
LLM-based agents have shown promising capabilities in a growing range of software engineering (SWE) tasks. However, advancing this field faces two critical challenges. First, high-…
Training Long-Context, Multi-Turn Software Engineering Agents with Reinforcement Learning
Alexander Golubev, Maria Trofimova, Sergei Polezhaev +9
Research on applications of reinforcement learning (RL) to large language models has mostly been focused on single-turn problems, such as mathematical reasoning or single-shot code…
Guided Search Strategies in Non-Serializable Environments with Applications to Software Engineering Agents
Karina Zainullina, Alexander Golubev, Maria Trofimova +9
Large language models (LLMs) have recently achieved remarkable results in complex multi-step tasks, such as mathematical reasoning and agentic software engineering. However, they o…