Accepted at ASE 2026 · Munich · Oct 12–16

Kozuchi Agent

A language-agnostic, open-weight agent for software repair — state-of-the-art parameter efficiency on SWE-bench Verified with a 27B model.

— published in ASE 2026, the 41st IEEE/ACM International Conference on Automated Software Engineering, Munich, Germany

0%SWE-bench Verified
(374 / 500 resolved)
#1open-weight agent
(rank 12 / 135 overall)
0Bparameters — smallest model
on the Pareto frontier
0%Multi-SWE-bench Java
(rank 4 / 42, #1 open-weight)
01 · Overview

Small model, frontier results

browse repository ↗

Kozuchi Agent is a phase-driven software-repair agent that pairs a 27B open-weight model (Qwen3.5-27B) with reuse orchestration, K-candidate trajectory sampling, and an agent-internal selector — no closed-source verifier and no hidden-test access at selection time. It resolves 74.80% of SWE-bench Verified and transfers to Java (32.03% on Multi-SWE-bench) without language-specific changes.

SWE-bench Verified success rate vs. model parameters; Kozuchi at 74.8% with 27B sits on the Pareto front
Kozuchi mini-swe-agent + Qwen3.5-27B reaches 74.8% on SWE-bench Verified with ~10–40× fewer parameters than peers at comparable success rates. — click to open in figures/
02 · The Paper

Read the paper

arXiv:2608.15579 ↗
First page of the Kozuchi Agent paper
Click the snapshot to open the full PDF
  • Preprint on arXiv: arXiv:2608.15579 — accepted at ASE 2026 Industry Showcase (13 pages, 4 figures).
  • ASE 2026 presentation: Wed 14 Oct 2026, 10:30–10:45 at Forum 2 (Industry Showcase).
  • ACM Digital Library: doi:10.1145/3832783.3834531.
  • Six research questions covering headline performance, peer comparison, variance analysis, failure taxonomy, test-time-scaling selection lift, and cross-language transfer.
  • Statistically grounded: Wilson CIs, repo-level cluster bootstrap, McNemar paired tests with BH-FDR correction across 17 open-weight peers.
  • Fully rebuildable: paper/build.sh compiles the PDF offline with bundled ACM class files.
  • Every cited number maps to a script via INDEX.md; every reference is annotated in REFERENCES_INDEX.md.
03 · Poster

ASE 2026 conference poster

download A0 PDF ↗
ASE 2026 A0 poster for Kozuchi Agent
Click the snapshot to open the full A0 PDF
  • A0 portrait (841 × 1189 mm) two-column poster for the ASE 2026 Industry Showcase in Munich.
  • Same claims as the paper: 27B open-weight Qwen3.5, no fine-tuning, no hidden-test access at selection, 74.8% on SWE-bench Verified.
  • QR on the poster points at this site — paper, artifact, code, and reproduction in one scan.
  • Print-ready PDF (~2.1 MB), generated from the editable PPTX in the paper repository.
04 · Slides

ASE 2026 talk deck

download PPTX ↗
Title slide of the ASE 2026 Kozuchi Agent talk
Click the snapshot to download the PPTX
  • Industry Showcase talk: From Phases to Patches — Wed 14 Oct 2026, 10:30–10:45 at Forum 2. Session page ↗
  • PowerPoint deck (From_Phases_to_Patches_ASE_2026_v6.pptx, ~4.5 MB) with extra backup slides after the 12-minute talk path.
  • Same Fujitsu template as the conference talk: phases, cross-agent tests, results, and takeaways.
05 · Architecture

How it works

browse figures/ ↗

Industrial software-engineering teams increasingly prefer LLM agents that turn bug reports into correct, test-passing patches — but operating such agents at industrial scale raises harness-level problems: long horizons, tool-use discipline, context persistence, multi-cluster execution, and evaluation reuse. Kozuchi Agent answers with a language-agnostic open-weight agent and a CI-operated evaluation pipeline built on five pillars:

Explicit phasesAn 8-phase / 15-edge finite-state workflow (reproduce → test synthesis → localization → fix → verify → report) with 14 phase-gated skills.
Persistent intermediate stateDurable shared memory — scripts, generated tests, traces, handovers, patch notes — survives across phases and candidates.
Deterministic SE toolsThree active, deterministic software-engineering tools instead of free-form shell improvisation.
Model-independent actionsA turn-level action normalizer keeps the action interface identical across backbone models.
Cross-agent test-time selectionK=8 candidate trajectories per instance, ranked by a K×K selector using only agent-chosen tests — never hidden benchmark tests.
Auditable & repeatableNo proprietary models, no fine-tuning, no large backbones — behavior is reproducible from a locally hosted Qwen3.5-27B.
Kozuchi Agent construction map: phase-driven agent, reuse orchestration, K candidate trajectories, selector
End-to-end construction map: hidden benchmark tests are used only by the official evaluator, never during selection. — click to open in figures/

The same harness transfers from Python to Java unchanged: per-phase scaffold behavior matches the Python run within ±5 percentage points on every phase, and the remaining failures are dominated by semantic correctness, Java-specific harness failures, and selection errors — not proprietary-model advantages.

06 · Results

Results at a glance

browse stats/ ↗
Python vs Java headline rates and per-phase message share
Cross-track transfer: one agent, two languages.
Per-repository resolution heatmap vs peers
Per-repository resolution vs. open-weight peers.
07 · Reproducibility

Reproduce every number

open reproduce.sh ↗

The artifact is self-contained: every numeric value, figure, and table in the paper is regenerable from files inside the repository, and INDEX.md maps each claim to its script, its paper section, and its regenerated output.

$ git clone https://github.com/marscod/kozuchi-agent-artifact
$ cd kozuchi-agent-artifact
$ ./reproduce.sh                # every paper number + figure (stats/, figures/, plots/)
$ ./reproduce.sh --paper        # also rebuild paper/main.pdf
$ ./decompress.sh               # optional: restore 22 GB of raw trajectories from data/archives/

Requirements: Python 3.10+ (scripts/numbers/ needs no third-party packages); matplotlib pandas seaborn numpy openpyxl for figures (auto-handled via uv when available); TeX Live only for --paper.

🗃 Dataset note: the Zenodo record (trajectory datasets for Python and Java) is under review and not yet publicly published. Until the record goes live, use the reviewer preview link and place the downloaded archives under data/archives/.
08 · Artifact

Artifact map

browse all files ↗

Every card opens the corresponding folder or file in the repository.

09 · Citation

If you build on this artifact, please cite the paper. The DOI is 10.1145/3832783.3834531 (ACM Digital Library). Session page: ASE 2026 Industry Showcase. A freely available preprint is on arXiv:2608.15579.

ACM Reference Format

Mehdi Bahrami, Kosaku Kimura, Satoshi Munakata, Satoshi Nakashima, Yu Ishikawa, Kosuke Maeda, Nao Soma, Kenichi Kobayashi, Keisuke Miyazaki, Keizo Kato, Shigeki Fukuta, Tatsuo Kumano, Nobutaka Imamura, Kevin Musgrave, Shahbaz Abdul Khader, Kwun Ho Ngan, Joe Townsend, Fayas Asharindavida, Matthieu Parizy, Akira Sakai, Yuma Ichikawa, Yang Zhao, Michiaki Takizawa, Taku Fukui, Hiroki Ohtsuji, Wei-Peng Chen, and Hiromichi Kobashi. 2026. Kozuchi Agent: A Language-Agnostic Open-Weight Agent for Software Repair. In Proceedings of the 41st IEEE/ACM International Conference on Automated Software Engineering (ASE '26), October 12–16, 2026, Munich, Germany. ACM, New York, NY, USA, 13 pages. https://doi.org/10.1145/3832783.3834531

BibTeX

@inproceedings{kozuchi2026, author = {Bahrami, Mehdi and Kimura, Kosaku and Munakata, Satoshi and Nakashima, Satoshi and Ishikawa, Yu and Maeda, Kosuke and Soma, Nao and Kobayashi, Kenichi and Miyazaki, Keisuke and Kato, Keizo and Fukuta, Shigeki and Kumano, Tatsuo and Imamura, Nobutaka and Musgrave, Kevin and Abdul Khader, Shahbaz and Ngan, Kwun Ho and Townsend, Joe and Asharindavida, Fayas and Parizy, Matthieu and Sakai, Akira and Ichikawa, Yuma and Zhao, Yang and Takizawa, Michiaki and Fukui, Taku and Ohtsuji, Hiroki and Chen, Wei-Peng and Kobashi, Hiromichi}, title = {Kozuchi Agent: A Language-Agnostic Open-Weight Agent for Software Repair}, year = {2026}, isbn = {979-8-4007-2882-2}, publisher = {Association for Computing Machinery}, address = {New York, NY, USA}, url = {https://doi.org/10.1145/3832783.3834531}, doi = {10.1145/3832783.3834531}, booktitle = {Proceedings of the 41st IEEE/ACM International Conference on Automated Software Engineering}, numpages = {13}, keywords = {AI agent, automated program repair, SWE-bench, experience report}, location = {Munich, Germany}, series = {ASE '26} }
10 · One-Minute Tour

The artifact in motion

Headline Pareto frontier, construction map, cross-language transfer, peer heatmap, the paper, poster, and talk slides.

Animated tour cycling through the key figures of the Kozuchi Agent artifact
Click to explore the full repository.