# FeatBench: Towards More Realistic Evaluation of Feature-level Code Generation

arXiv: https://arxiv.org/abs/2509.22237 · HTML: https://mlsanity.com/paper/2509.22237/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 14 of its 22 references that are in the map.
- First posted on arXiv 26 Sep 2025.

## Builds on (most cited first, 14 of 14)

- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Code Llama: Open Foundation Models for Code](https://mlsanity.com/paper/2308.12950/) (2023)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://mlsanity.com/paper/2405.15793/) (2024)
- [Datasets: A Community Library for Natural Language Processing](https://mlsanity.com/paper/2109.02846/) (2021)
- [Agentless: Demystifying LLM-based Software Engineering Agents](https://mlsanity.com/paper/2407.01489/) (2024)
- [ClassEval: A Manually-Crafted Benchmark for Evaluating LLMs on Class-level Code Generation](https://mlsanity.com/paper/2308.01861/) (2023)
- [SWE-PolyBench: A multi-language benchmark for repository level evaluation of coding agents](https://mlsanity.com/paper/2504.08703/) (2025)
- [EvoCodeBench: An Evolving Code Generation Benchmark Aligned with Real-World Code Repositories](https://mlsanity.com/paper/2404.00599/) (2024)
- [DevEval: A Manually-Annotated Code Generation Benchmark Aligned with Real-World Code Repositories](https://mlsanity.com/paper/2405.19856/) (2024)
- [Seed-Coder: Let the Code Model Curate Data for Itself](https://mlsanity.com/paper/2506.03524/) (2025)
- [Doc2Feat-Bench: Evaluating Documentation-Driven Feature Addition](https://mlsanity.com/paper/2507.18130/) (2025)

## Built on by (most cited first, 6 of 6)

- [LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering](https://mlsanity.com/paper/2608.28281/) (2026)
- [Chasing the Public Score: User Pressure and Evaluation Exploitation in Coding Agent Workflows](https://mlsanity.com/paper/2604.20200/) (2026)
- [AlphaEval: Evaluating Agents in Production](https://mlsanity.com/paper/2604.12162/) (2026)
- [SWE-WebDevBench: Evaluating Coding Agent Application Platforms as Virtual Software Agencies](https://mlsanity.com/paper/2605.04637/) (2026)
- [FeatLens: Feature-Guided Dynamic Code Graph Construction and Retrieval for Repository-Level Code Generation](https://mlsanity.com/paper/2609.26480/) (2026)
- [SWE-Journey: Towards More Realistic Evaluation of Coding Assistants through Long-Horizon, Multi-Turn Interaction](https://mlsanity.com/paper/2610.11559/) (2026)

## Abstract (arXiv)

Evaluating Large Language Models (LLMs) on repository-level feature implementation is a critical frontier in software engineering. However, establishing a benchmark that faithfully mirrors realistic development scenarios remains a significant challenge. Existing feature-level benchmarks generally suffer from two primary limitations: unrealistic task inputs enriched with code hints and significant data leakage risks due to their static nature. To address these limitations, we propose a new benchmark - FeatBench, which introduces the following advances: (1) Realistic Task Inputs. Task inputs consist solely of natural language requirements, strictly devoid of code hints (e.g., function signatures). This format mirrors realistic software development by requiring agents to independently bridge the gap between abstract user intent and concrete code changes. (2) Evolving Data. FeatBench employs a fully automated pipeline to construct new benchmark versions from the latest repositories, effectively mitigating data contamination. The initial release comprises 157 tasks sourced from 27 actively maintained repositories. We evaluate two state-of-the-art agent frameworks with four leading LLMs on FeatBench. The results reveal that FeatBench poses a significant challenge, with the highest resolved rate reaching only 29.94%. Crucially, our analysis uncovers a prevalent behavioral pattern of aggressive implementation, which leads to "scope creep" and widespread regressions where agents break existing features by diverging from the user's explicit intent. We release FeatBench, our automated pipeline, and all experimental results to facilitate further community research.