Composer 2 Technical Report

• URL: http://arxiv.org/abs/2603.24477v2
• Date: Thu, 26 Mar 2026 01:57:05 GMT
• Title: Composer 2 Technical Report
• Abstract summary: Composer 2 is a specialized model designed for agentic software engineering.<n>The model is trained in two phases: first, continued pretraining to improve the model's knowledge and latent coding ability.<n>We develop infrastructure to support training in the same Cursor harness that is used by the deployed model.
• Score: 93.84516486051359
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Composer 2 is a specialized model designed for agentic software engineering. The model demonstrates strong long-term planning and coding intelligence while maintaining the ability to efficiently solve problems for interactive use. The model is trained in two phases: first, continued pretraining to improve the model's knowledge and latent coding ability, followed by large-scale reinforcement learning to improve end-to-end coding performance through stronger reasoning, accurate multi-step execution, and coherence on long-horizon realistic coding problems. We develop infrastructure to support training in the same Cursor harness that is used by the deployed model, with equivalent tools and structure, and use environments that match real problems closely. To measure the ability of the model on increasingly difficult tasks, we introduce a benchmark derived from real software engineering problems in large codebases including our own. Composer 2 is a frontier-level coding model and demonstrates a process for training strong domain-specialized models. On our CursorBench evaluations the model achieves a major improvement in accuracy compared to previous Composer models (61.3). On public benchmarks the model scores 61.7 on Terminal-Bench and 73.7 on SWE-bench Multilingual in our harness, comparable to state-of-the-art systems.

Related papers

• On-the-Fly Input Adaptation for Reliable Code Intelligence [11.136449698197174]
  Code language models (CLMs) play a central role in software engineering across both generation and classification tasks.<n>Existing solutions address this by retraining the model, modifying its architecture, or re-engineering prompts.<n>This work proposes an alternative strategy based on on-the-fly input adaptation, which improves model behavior without altering its parameters.
  arXiv  Detail & Related papers  (2026-05-19T04:55:49Z)
• IQuest-Coder-V1 Technical Report [56.045533451719905]
  We introduce the IQuest-Coder-V1 series-(7B/14B/40B/40B-Loop), a new family of code large language models (LLMs)<n>IQuest-Coder-V1 achieves state-of-the-art performance among competitive models across critical dimensions of code intelligence.
  arXiv  Detail & Related papers  (2026-03-17T16:15:31Z)
• Youtu-LLM: Unlocking the Native Agentic Potential for Lightweight Large Language Models [78.73992315826035]
  We introduce Youtu-LLM, a lightweight language model that harmonizes high computational efficiency with native agentic intelligence.<n>Youtu-LLM is pre-trained from scratch to systematically cultivate reasoning and planning capabilities.
  arXiv  Detail & Related papers  (2025-12-31T04:25:11Z)
• AgentMath: Empowering Mathematical Reasoning for Large Language Models via Tool-Augmented Agent [80.83250816918861]
  Large Reasoning Models (LRMs) like o3 and DeepSeek-R1 have achieved remarkable progress in natural language reasoning with long chain-of-thought.<n>However, they remain computationally inefficient and struggle with accuracy when solving problems requiring complex mathematical operations.<n>We present AgentMath, an agent framework that seamlessly integrates language models' reasoning capabilities with code interpreters' computational precision.
  arXiv  Detail & Related papers  (2025-12-23T19:57:49Z)
• Does In-IDE Calibration of Large Language Models work at Scale? [4.707628898226459]
  Post-hoc calibration of internal model confidences aims to align probabilities with an acceptability measure.<n>We develop a scalable and flexible calibration framework which can be used to obtain calibration weights for open-source models.<n>Through a large-scale analysis of over 24 million real-world developer interactions, we find that a general, post-hoc calibration model based on Platt-scaling does not, on average, improve the reliability of model confidence signals.
  arXiv  Detail & Related papers  (2025-10-26T10:15:03Z)
• Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models [49.911784762244814]
  TraceRL is a trajectory-aware reinforcement learning framework for diffusion language models (DLMs)<n>We derive a series of state-of-the-art diffusion language models, namely TraDo.<n>TraDo-8B-Instruct achieves relative accuracy improvements of 6.1% over Qwen2.5-7B-Instruct and 51.3% over Llama3.1-8B-Instruct on mathematical reasoning benchmarks.
  arXiv  Detail & Related papers  (2025-09-08T17:58:06Z)
• Scaling Reasoning without Attention [44.42046576158219]
  We introduce ourmodel, an attention-free language model that addresses both issues through architectural and data-centric innovations.<n>Our model eliminates the need for self-attention and key-value caching, enabling fixed-memory, constant-time inference.<n>On benchmark evaluations, ourmodel-7B outperforms strong Transformer and hybrid models of comparable scale.
  arXiv  Detail & Related papers  (2025-05-28T14:52:15Z)
• Verbal Process Supervision Elicits Better Coding Agents [0.9558392439655016]
  This work introduces CURA, a code understanding and reasoning agent system enhanced with verbal process supervision (VPS)<n>CURA achieves a 3.65% improvement over baseline models on challenging benchmarks like BigCodeBench.
  arXiv  Detail & Related papers  (2025-03-24T09:48:59Z)
• 360Brew: A Decoder-only Foundation Model for Personalized Ranking and Recommendation [18.330496007364882]
  We introduce our research pre-production model, 360Brew V1.0, a 150B parameter, decoder-only model that has been trained and fine-tuned on LinkedIn's data and tasks.<n>This model is capable of solving over 30 predictive tasks across various segments of the LinkedIn platform, achieving performance levels comparable to or exceeding those of current production systems.
  arXiv  Detail & Related papers  (2025-01-27T19:14:52Z)
• Code Representation Learning At Scale [75.04686476303436]
  We fuel code representation learning with a vast amount of code data via a two-stage pretraining scheme. We first train the encoders via a mix that leverages both randomness in masking language modeling and the structure aspect of programming language. We then enhance the representations via contrastive learning with hard negative and hard positive constructed in an unsupervised manner.
  arXiv  Detail & Related papers  (2024-02-02T22:19:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.