SWE-Pruner Pro: The Coder LLM Already Knows What to Prune

Hugging Face Daily Papers Papers
code-generation context-pruning llm-agents swe-bench efficiency

Summary

SWE-Pruner Pro leverages the coding agent's own internal representations to prune long code context, saving up to 39% of tokens while maintaining or improving task performance on multi-turn benchmarks.

Pruning long context for coding agents has been a vital technology for efficient context management. While existing context pruning methods such as SWE-Pruner realize this by attaching a separate code classifier, we find the agent itself encodes internal representations indicating the relevance of code context when reading tool output. Based on this finding, we propose SWE-Pruner Pro, which prunes tool outputs directly inside the agent. Concretely, a small head turns the agent's own internal representations into a keep-or-prune label for each line, with a length-aware embedding keyed to each tool output's line count. Across two open-weight backbones and four multi-turn benchmarks, SWE-Pruner Pro saves up to 39% of prompt and completion tokens while preserving task quality, with bounded inference overhead. Notably, on MiMo-V2-Flash SWE-Pruner Pro additionally raises the SWE-Bench Verified resolve rate by +3.8% and the long-context Oolong accuracy by +2.2 points.
Original Article
View Cached Full Text

Cached at: 07/21/26, 06:35 AM

Paper page - SWE-Pruner Pro: The Coder LLM Already Knows What to Prune

Source: https://huggingface.co/papers/2607.18213

Abstract

Pruninglongcontextforcodingagentshasbeenavitaltechnologyforefficientcontextmanagement.WhileexistingcontextpruningmethodssuchasSWE-Prunerrealizethisbyattachingaseparatecodeclassifier,wefindtheagentitselfencodesinternalrepresentationsindicatingtherelevanceofcodecontextwhenreadingtooloutput.Basedonthisfinding,weproposeSWE-PrunerPro,whichprunestooloutputsdirectlyinsidetheagent.Concretely,asmallheadturnstheagent’sowninternalrepresentationsintoakeep-or-prunelabelforeachline,withalength-awareembeddingkeyedtoeachtooloutput’slinecount.Acrosstwoopen-weightbackbonesandfourmulti-turnbenchmarks,SWE-PrunerProsavesupto39%ofpromptandcompletiontokenswhilepreservingtaskquality,withboundedinferenceoverhead.Notably,onMiMo-V2-FlashSWE-PrunerProadditionallyraisestheSWE-BenchVerifiedresolverateby+3.8%andthelong-contextOolongaccuracyby+2.2points.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2607\.18213

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2607.18213 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.18213 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.18213 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Context Pruning for Coding Agents via Multi-Rubric Latent Reasoning

arXiv cs.AI

LaMR introduces a structured pruning framework for coding agents that decomposes code relevance into semantic evidence and dependency support dimensions, using dedicated CRFs and a mixture-of-experts gate to reduce token usage by up to 31% while maintaining or improving task performance.

Cut Your Losses! Learning to Prune Paths Early for Efficient Parallel Reasoning

Hugging Face Daily Papers

This paper introduces STOP (Super Token for Pruning), a lightweight method that learns to prune unpromising reasoning paths early during parallel decoding by appending learnable tokens and reading KV cache states, achieving 70% token reduction while improving performance on AIME and GPQA benchmarks.