Large language model (LLM) agents increasingly operate over evolving software repositories, where success depends on repository-specific procedural knowledge: which APIs to call, which scripts to run, and which conventions the current release expects. Agent skills externalize this knowledge into reusable units, and prior work shows that they can improve agent performance. What remains unclear is whether that improvement is durable. The same version specificity that makes a skill useful also makes it fragile: after a release, it may become stale without raising any explicit signal, while continuing to provide obsolete guidance. Externalizing knowledge into a skill can therefore make its decay invisible. We study whether agents can keep this externalized knowledge current. Repo2Skill-Evo casts each release transition as a skill-maintenance task: given a V1 skill set and the official V1-to-V2 patch, an agent must update obsolete skill content while preserving guidance that remains valid. Across 57 real-world repositories and 105 selected release transitions, every evaluated transition invalidates part of the V1 skill set. Yet six frontier agents reach only 29.9%-69.7% avg@3 macro F1 under a patch-grounded removal metric that balances stale-content recall against over-editing precision. Across runs, two opposing errors dominate: incomplete coverage of affected files in the skill set leaves stale content untouched, while overbroad editing is associated with higher recall but lower precision. Repository skills go stale in silence, and even frontier agents cannot reliably maintain them.
Large Language Models (LLMs) are moving from code completion toward repository-scale agents that retrieve context, edit files, execute tools, and participate in security-sensitive workflows. The evidence for these systems, however, remains divided between software engineering evaluations centered on functional task completion and software security evaluations centered on vulnerability detection, secure generation, or exploit-oriented validation. This evidence-centered structured survey synthesizes representative work available through May 31, 2026 across software engineering tasks, software security tasks, adaptation mechanisms, artifact granularity, and evaluation design. In addition to a task taxonomy, we introduce an assurance framework that separates functional correctness, security, operational reliability, evidence provenance, and agent authority. The review shows that execution feedback and repository access can substantially improve engineering task completion, but do not by themselves establish security; conversely, static-analysis labels or vulnerability-classification scores rarely establish deployable correctness. We identify recurring validity threats--weak test oracles, duplicated and temporally leaked data, changing agent harnesses, proxy-only security checks, and under-reported budgets and human intervention--and derive a minimum reporting protocol for cross-study comparison. The resulting research agenda prioritizes jointly secure-and-functional benchmarks, repository-scale threat models, calibrated human oversight, longitudinal maintainability evidence, and reproducible agent evaluation. The central conclusion is that model capability should be judged as an assurance case supported by task-appropriate evidence, rather than by a single benchmark score.
Mehdi Bahrami, Kosaku Kimura, Satoshi Munakata +24cs.SE cs.AI cs.ET cs.PL
Industrial software-engineering teams increasingly need LLM agents that turn bug reports into correct patches, yet benchmark-scale operation adds long horizons, tool-use discipline, context persistence, heterogeneous clusters, and evaluation reuse. We present Kozuchi Agent, a language-agnostic open-weight repair agent and CI-operated evaluation pipeline. Explicit phases, persistent state, deterministic tools, a model-independent action interface, and cross-agent test-time selection make runs auditable and repeatable. With locally hosted Qwen3.5-27B, no fine-tuning, and TTS@8, Kozuchi resolves 374/500 SWE-bench Verified instances on the official evaluator. Unchanged on Multi-SWE-bench Java, the same 27-billion-parameter agent resolves 41/128 instances (32.03%), ranking first among strict open-weight submissions and fourth of 42 overall; on Python it ranks 12th of 135 and first among open-weight systems. Per-phase behavior remains within +/-5 percentage points across languages. Remaining failures mainly reflect semantic correctness, Java-specific harness issues, and selection errors. Across both tracks, results compare favorably with open/local peers by parameter count. Analysis of candidate diversity, selector regret, and patch reliability shows that the remaining gap is primarily semantic correctness and selection rather than edit formatting or proprietary-model access. Operationally, reusable CI stages reduce operator touch-points from five to one across heterogeneous internal clusters.
LLM coding agents issue Bash commands through interfaces that may serialize, wrap, and reparse model output. Matched execution scores alone cannot distinguish command-generation errors from failures introduced after generation. QuoteBench measures this boundary with exact final-state validation on 56 one-shot tasks from 14 incident-derived families, crossing the generation contract with the execution transport around one deliberately unescaped added parser. Escaping at the interpolation point reproduces each replayed reply's raw-path outcome, so any recovery under a disclosed boundary must come from the model changing its generation. Across eight same-window configurations, replaying the same reply through the added parser lowers success by 55.4 to 73.2 percentage points; disclosure recovers 30.4 to 60.7 points for six configurations, and zero or slightly negative for the other two. Raw generation is nearly saturated at the frontier; boundary adaptation is what still separates models. GPT-5.6-sol's matched gap of -3.6 points hides -64.3 points of damage and +60.7 points of compensation. The deployment configuration reorders models: one reversal among 26 comparable pairs is unambiguous and four more sit on single-task margins. Evaluations of command-issuing agents should report the model configuration, generation contract, execution path, operating point, and final-state validator rather than treat a matched score as an intrinsic model property.
Code agents spend much of their effort simply locating the right code inside a repository. Two approaches dominate current practice. In Semantic Search, the agent retrieves code blocks from a vector index built from the repository in advance. In Deep Agentic Search (also known as grep-search by subagent), a planning agent delegates the exploration to a separate subagent that works in an isolated context window and returns only a condensed result. The second design, which is considered good context engineering practice, exists to protect the main agent from context pollution (also known as context rot), the loss of accuracy that occurs as unrelated material accumulates in the context window. Recent code agents (such as Claude Code, Codex, Antigravity, etc) have adopted it quickly, but there is little evidence on whether it produces better answers. We compare the two approaches on SWE-QA, a benchmark for repository-level code question answering. Semantic search answered 65.2% of questions correctly against 46.2% for deep agentic search, and it produced each correct answer at less than half the cost. To explain the gap, we then coded every failed run into a taxonomy of failure modes. The taxonomy shows that deep agentic search did not remove failures but introduced a new class of them: the single largest share of its failures, 41.8%, occurred at the hand-off between the planner and its sub-agent, and these were usually silent, ending in a fluent and confident answer that was wrong. Deep agentic search addresses a real problem and is now the preferred design in many code agents. However, our results show that the protection it offers may not be free, and that for read-only questions over a repository that can be indexed, retrieval was the stronger and cheaper option.
Dotan Davidovich, Yair Amar, Hai Rozencwajg +1cs.CR cs.AI
Coding agents increasingly run inside organizations whose security controls (scoped credentials, restricted egress, read-only filesystems, non-root execution) constrain them like any other software. Existing benchmarks, however, evaluate agents almost exclusively in permissive sandboxes, so it is unknown how performance changes when policy is enforced. In this work, we evaluate 12 coding agents on Terminal-Bench 2.1 across nested security policy levels derived from common real-world enterprise restrictions. Hardening is never free but far from uniform: under the strictest policy, success losses reach 18.3 points and cost inflation 167.3\%, and the two axes disagree; the model that best preserves success is also the one that loses the most efficiency, so model choice is policy-dependent. Beyond aggregate scores, we characterize how agents behave when policy blocks their actions and decompose the failures hardening induces: runs grind into timeouts or wrong solutions rather than stopping early, in a mix that differs by model. To ground comparisons, we verify task solvability under the strictest policy, separating model failures from tasks the policy forecloses. We release Boundary-Bench, an open-source hardening plugin enabling policy-constrained evaluation of coding agents on Terminal-Bench and compatible benchmarks.
Open source communities have been flooded with AI-generated contributions. In defense, they have written contribution rules to regulate coding agents' behavior, spanning from a total ban, mandatory disclosure, to verification gates and human sign-offs. Yet, whether coding agents read and follow those rules, and behave in open source repositories, remains unknown. To estimate real-world rule compliance of coding agents, we curate 106 issues from 49 repositories containing AI contribution rules into RepoComplianceBench. We judge the trajectory of each run against the repository's rules, measuring whether the agent refuses to contribute, discloses its assistance truthfully, clears the required verification gates, or escalates critical steps to a human. We also test if extra prompts, rule disclosure, or feedback from the compliance verifier help with the situation. Our experiments on four frontier models show that today's agents almost never proactively retrieve the contribution rules. Agents pick up disclosure and verification with reminder prompts, rule quotes, and verifier feedback; however, they never refuse to contribute in AI-banned repositories under any condition we tested. The status reveals that verification and disclosure issues are solvable with existing mechanisms, yet enforcing bans and human escalations remains an open problem.
Persistent context files (AGENTS.md, CLAUDE.md) are standard practice for guiding AI coding agents, yet evidence for their effectiveness is contradictory. We present a controlled ablation of context-injection strategy across two frontier agents (Claude Code and Codex), 17 real tasks from 3 repositories (15 shared + 2 Codex-only), and 288 evaluated runs with gold-test evaluation. Context strategy does not measurably move correctness on either agent (bounded to <=10-15pp via equivalence testing). A failure-mode triage reveals why: agents fail on implementation skill---feature design, pattern selection, exact wiring---not missing repository knowledge that a context file could supply; a manipulation probe confirms the real AGENTS.md never converts a near-miss to a pass on either agent. We further show that borderline task difficulty is agent-specific (Spearman rho=0.75), offering a candidate explanation for prior contradictions: single-agent studies draw tasks from different agents' informative bands. We release all code, data, and analysis.
Generate--test--revise loops are common in coding agents, but repetition alone provides no reliability guarantee. We study the gap between finding a correct patch and retaining, verifying, and submitting it. A sealed five-seed study over 30 HumanEval repairs produces 900 three-revision trajectories. Under forced revision, current correctness with current traces falls from 0.820 after one revision to 0.673 after two, although ever-correct rises to 0.847. Two common-state studies use 2,430 branches from identical frozen programs to remove post-treatment risk-set bias. In a prespecified 14B replication, stale traces harm 34/135 correct starts versus 4/135 with current traces, a 22.2-point increase (task-cluster 95\% CI $[8.9,37.0]$, exact Holm $p=0.0337$). A prospective 540-rollout policy eliminates observed correct-start harm but reduces wrong-start repair and fails its joint criterion. Repository experiments over 24 bugs and four coder stacks expose floor effects and component heterogeneity without Holm-significant effects. We therefore separate admission, preservation, grounded certification, competence, and liveness. We derive an evidence-bound typed loop contract and instantiate its mechanically enforceable subset in a reference implementation that binds verifier evidence to exact code states, preserves verified checkpoints, and emits auditable admission receipts. The implementation is an executable specification and conformance artifact, not evidence of improved repair competence or calibrated verifier dependence.
The recent emergence of vibe-coding workflows is changing what coding agents are expected to do. Instead of merely completing code under fully specified instructions, agents are increasingly expected to transform incomplete product intent into working software by combining various abilities including planning, requirement clarification, tool use, debugging, and repository-level construction. Yet existing benchmarks have not fully caught up with this shift, evaluating agents on static, fully specified tasks. In this paper, we introduce ICAE-Bench, a benchmark for evaluating coding agents under interactive project-building settings. The basic idea is to start from a fuzzy product requirement, simulating the dynamic paradigm with an automated User Agent. To make this setting both realistic and evaluable, ICAE-Bench introduces three key designs. First, to avoid the ambiguity of unconstrained fuzzy requirements, each task derives ambiguity from a precise real open-source repository with executable behavior. Second, to ensure high-quality and reproducible user simulation, ICAE-Bench grounds interaction through User Agent Data, allowing the User Agent to reveal hidden constraints without inventing new requirements or leaking implementation artifacts. Third, to evaluate open-ended repositories fairly, ICAE-Bench uses standardized black-box tests together with multi-dimensional diagnostics, including functional correctness, semantic and API similarity, structural fidelity, design quality, and interaction quality.
Nursultan Askarbekuly, Mohamad Al Mdfaa, Ahmed Helaly +2cs.SE cs.AI
Coding agents can now be left alone to improve software against a score. In this pattern--recently popularized as "autoresearch"--the agent receives a dataset, an evaluation script, and one editable file, and iterates without supervision: modify the code, measure, keep the change if the score improves. But what does the agent actually optimize--the developer's intent, or the literal number? We ran this loop on a real production task: deciding which Quranic verses appear in a noisy speech-recognition transcript and splitting the transcript by verse. Two frontier coding agents, Claude Code and OpenAI Codex, started from the same blank file with the same instructions, budget, and reasoning effort, three runs each. Both independently invented the same algorithm (canonicalization, n-gram anchoring, dynamic-programming alignment)--and then diverged. Claude stopped early with compact, general code. Codex drove the score ~10x lower, largely by memorizing answers to individual evaluation rows (19-41 hardcoded verse ids per run): a clean natural instance of specification gaming by a production agent. In a preregistered second study, we added a held-out test set and told both agents it existed. The memorization vanished, and the score gap vanished with it--yet Codex's general core transferred better and more consistently (held-out detection+split 0.085+/-0.004 vs. 0.121+/-0.031), losing only on one missed rejection of non-recitation input. Two exploratory community arms (Cursor, Antigravity) are consistent with the pattern. Every agent's held-out solution matched or beat the hand-engineered pipeline it was built to replace--the best by an order of magnitude--and now runs in production. From the ways agents exploited our harness--reading sibling runs through shared git state, leaving notes to "future runs" in persistent memory--we distill five design rules for evaluating autonomous agents.
Autonomous coding agents increasingly execute multi-step software work, but lifecycle states such as reviewed, tested, DONE, and ready-to-merge remain claims unless supported by current evidence. We present Proof-or-Stop Lifecycle Control, a method that permits lifecycle transitions only when fresh, tracked-source-state-bound, mechanically verifiable evidence satisfies the relevant gate. The method treats agent outputs as claims rather than lifecycle state, and uses proof operationally to mean gate-admissible evidence under a stated trust model, not semantic program correctness. We evaluate an open-source implementation through mechanism tests, a powered control-policy ablation, and operated self-application evidence. The unattended-loop engine passed 10 of 10 scenarios with zero false-DONE, and local-key receipt bundles rejected 18 tamper classes with zero false accepts. In a 9,240-cell ablation, the pre-registered A4 versus A2-prime comparison reduced visible-pass/hidden-fail amplification from 31 of 1,800 injected cells under a compute-budgeted naive loop to 2 of 1,800 under the gated loop, a 1.6 percentage-point improvement in not-amplified rate with a 95 percent confidence interval of [0.8, 2.5]. A near-compute A3 versus A4 comparison, 14 of 1,800 versus 2 of 1,800, indicates that the gain is associated with enforcing review as a lifecycle gate rather than merely adding a reviewer. The self-application corpus contains 565 stories and 1,007 review findings, with 94.8 percent resolved, plus a 68-row high/critical cross-vendor exhibit. These results support Proof-or-Stop as a model-agnostic, host-neutral control layer for deciding which autonomous-agent claims a lifecycle may act on. The evaluation is limited to one model family, 24 ablation tasks, and a self-hosted corpus.
Large language models are rapidly moving towards closing the development cycle, transitioning from simple assistive companions to autonomous contributors deeply embedded into collaborative development environments. Despite their accelerated adoption, existing evaluation techniques are limited due to their fragmented nature and distorted projection of true model capabilities, often obtained from hypothetical syntactic scenarios. This research aims to bridge this gap by providing a comprehensive evaluation methodology for LLM-powered agents that is grounded in real-world software development practice. Our evaluation approach focuses on contamination-awareness, in-the-wild agentic behavior assessment, and trajectory-aware benchmarks and metrics capturing realistic coding contexts, human-aligned behavior, and model failure modes.
Compiler missed optimizations refer to cases in which compilers failed to optimize certain code. It takes many compiler developers' efforts to implement or patch such missed optimizations. In this paper, we present a systematic study of how well agents patch compiler missed optimizations. We identify a significant challenge that patching a missed optimization requires more than just fixing the reported case, and instead requires generalizing to similar cases. We construct a benchmark of real-world LLVM missed optimization issues and compare agent-generated patches with patches from developers in terms of optimization scope. Our results show that coding agents often optimize the given examples, but many generated patches either cover only part of the developer-intended scope or partially overlap with it; in some cases, they further generalize beyond the reference patch. We further introduce historical-knowledge augmentation techniques that leverage prior LLVM optimization pull requests through retrieval and distillation, showing that they improve developer-aligned generalization and yield practical benefits when applied to real-world IR.
We introduce SWE-Interact, a new testbed for evaluating coding agents on multi-turn, interactive, user-driven software engineering tasks. Existing frontier SWE benchmarks typically provide complete requirements upfront and evaluate agents on autonomous implementation. In contrast, SWE-Interact places agents in a realistic developer workflow: a carefully designed user simulator starts with vague or incomplete instructions, progressively reveals requirements, inspects the agent's workspace, and provides targeted feedback, revisions, and new constraints until the full task goal has been handed off. Grounded in large-scale studies of real coding-agent interactions, this setup tests whether agents can discover user intent, adapt to evolving requirements, and build on their own prior work. Across a suite of frontier and open-weight models, we find that strong performance on single-turn SWE tasks does not reliably transfer to multi-turn, user-driven workflows: the best-performing models solve roughly 50% of single-turn baseline tasks but only 25% of the corresponding SWE-Interact tasks. The strongest models in our evaluation, including Opus 4.8 and GPT 5.5, start strong even in the face of vague initial instructions, persevere until all the requirements are surfaced by the user, integrate them better and write clean code. However, they still suffer from over-agentic coding, forgetting requirements and technical mistakes. Weaker models start poorly under ambiguity, give up early, forget or ignore instructions and rework their code more. Overall, SWE-Interact measures an orthogonal, real-world capability axis for frontier model development: interactive goal discovery and iterative refinement with a user in the loop.
We introduce StaminaBench, a benchmark that measures the stamina of coding agents: how many consecutive interaction turns (change requests) they can handle before failing. Unlike the prevailing fraction-of-tasks-solved metric, this matches real vibe-coding where sessions run dozens or hundreds of turns. In StaminaBench, agents implement a REST API server and modify it across a tunable number of procedurally generated follow-up change requests - 100 in our experiments, resulting in codebases of up to 6,000 lines. Tests are generated fully programmatically without LLM involvement, ensuring reproducibility and reliability; change sequences are drawn from either a hardcoded or LLM-driven sampler, both constrained to a structured action space to ensure changes are valid. The agent and the server run in an isolated environment and communicate with the benchmark through HTTP, making testing fully black-box and language-agnostic. We evaluate six agent harnesses paired with seven open-source LLMs across 20 scenarios of 100 turns each and find that: (1) all the tested models fail within 5-6 turns, confirming that vibe-coding-style programming without thorough testing produces bugs; (2) passing test feedback back to the agent and allowing it to retry improves passed turn count by up to 12x; and (3) a good harness is required for strong performance: stronger models exhibit up to a 6x gap between their best and worst harness, while weaker models fail with any harness. We release the benchmark and the generated tasks to enable further research into multi-turn coding agent behavior. Benchmark code and data: github.com/amazon-science/StaminaBench.
Shanda Li, Qiuhong Anna Wei, Jingwu Tang +5cs.CL cs.AI cs.LG
Reproducing research results from papers and released code is central to scientific progress. Existing works have introduced benchmarks to evaluate whether LLM agents can assist with reproducibility, but they are difficult to scale due to their reliance on substantial manual effort for data curation and evaluation. We introduce ReproRepo, a scalable framework for reproducibility evaluation that leverages human-raised GitHub issues as naturally occurring supervision on realistic reproduction blockers. We instantiate ReproRepo on 1,149 recent machine learning papers from major conferences and evaluate four frontier model-agent configurations. Our results show that LLM agents, even without executing code, can identify many real-world reproducibility problems from paper-repository pairs: the best agent in our study, namely Codex with GPT-5.5, surfaces at least one semantically related human-reported blocker for ~90% of papers in the study. Further analysis shows that agents are particularly effective for surfacing visible failures and identifying the right semantic region, but may still be insufficient in exact localization. ReproRepo can serve as a reusable, scalable framework for future evaluations of LLM agents on real-world reproducibility auditing. Our code is released at https://github.com/LithiumDA/ReproRepo.
Game generation is an emerging application of coding agents, requiring models to transform natural-language specifications into playable interactive systems. Unlike traditional coding tasks, game generation takes place within a game engine, where scripts, scenes, assets, rendering, and runtime interactions must jointly produce coherent gameplay. We formalize end-to-end game generation as the problem of producing a complete game artifact that realizes a specification through observable player-game interaction in a target environment. We argue that evaluating this setting requires three desiderata: Engine Grounding, Artifact Completeness, and Interactive Verification. We propose an interaction-grounded evaluation framework that assesses executable gameplay through replayed demonstrations and rubric-guided multimodal judging. We instantiate this framework as GameCraft-Bench, a benchmark comprising 140 Godot tasks across 15 game families. Evaluations of frontier coding agents show that end-to-end game generation remains highly challenging: the strongest agent achieves only 41.46%, and most agents score below 40%. Further analysis reveals that while agents often implement recognizable mechanics, they struggle to deliver complete games with sufficient content, functional visual feedback, and coherent presentation. See https://tongxuluo.github.io/gamecraft-bench-website for demos, code, and data.
AI coding agents increasingly accept assigned software tasks, modify repositories under bounded authority, and return work packages for review. Prior work proposed the software delegation contract, covering the task, authority, returned work package, and acceptance context, as the unit of analysis for delegated coding work, but did not measure its effects. This paper reports a controlled pilot study of explicit delegation contracts for coding agents. We built a dependency-free TypeScript API task environment with seeded defects and documentation gaps, authored ten tasks across five families, and ran 64 agent executions across two model tiers under three conditions: a realistic issue-style prompt, an explicit delegation contract, and a contract with a required evidence bundle. Each run was scored with hidden acceptance tests, mutation checks, and scope analysis, then reviewed by three independent condition-blinded model-based reviewers using a fixed rubric, for 192 reviews. Explicit contracts did not improve objective task outcomes: all 64 runs passed hidden acceptance checks, with zero scope violations. They did improve reviewability. Evidence sufficiency improved in 22 of 30 paired comparisons and worsened in none (+0.83 on a 5-point scale, p < 0.0001, Cliff's delta = 0.66); reviewer ambiguity decreased (p = 0.035); changed-file lists, known-limitations sections, residual-risk sections, and reviewer checklists appeared mostly or only when demanded by the contract. Contracts cost +13% agent tokens and +38% wall-clock time, with larger effects for the weaker model tier. On these small tasks, delegation contracts bought reviewability rather than correctness.
LLM-based coding agents are usually evaluated in familiar software settings: mainstream languages, common libraries, and public repositories. These benchmarks remain important, but they can hide how agents behave when the language itself is unfamiliar. We evaluate six contemporary coding agents on four esoteric programming languages using a sequential setup with file editing, local execution, and hidden-test grading. Our protocol exposes capability differences between these agents that mainstream coding and agentic benchmarks such as SWE-Bench Verified and Terminal-Bench 2.0 compress into much narrower bands. We observe that the strongest agents, Claude Opus 4.6 and GPT-5.4 xhigh, often avoid writing the target language directly. On Brainfuck and Befunge-98, they write Python programs that generate target-language code and debug those generators locally. Forbidding this metaprogramming strategy causes large performance drops. Text guidance distilled from this strategy does not materially improve weaker agents. In contrast, Opus-derived Python helper code for building generators, with no solved benchmark programs or hidden-test answers, sharply improves Sonnet 4.6 and GPT-5.4 mini on the same problems, while Haiku 4.5 remains low. More interpreter calls and output tokens improve stronger agents but leave weaker agents near their original performance, indicating that these resources amplify useful strategies rather than create them. Together, these results show that strong coding agents adapt to unfamiliar languages by using tools, feedback, and workspace state to build a working model of the target language. Metaprogramming is the clearest case, but the broader gap is constructing and debugging a strategy that works under the target language's rules.
Thanawat Lodkaew, Johannes Ackermann, Soichiro Nishimori +3cs.LG cs.AI cs.CL stat.ME
A growing failure mode in agent evaluation and training is that models can achieve high evaluation scores by exploiting shortcuts instead of solving the intended task, producing deceptive performance. This makes evaluation scores unreliable as measures of true task-solving ability. We propose CapCode, a framework for constructing coding datasets with randomized tests whose best achievable non-cheating performance is deliberately capped below one. This capped-performance design gives evaluation scores a clearer interpretation: scores substantially above the cap are implausible and therefore provide evidence of cheating. To prevent cheating, we propose CapReward, a reward design based on the CapCode principle to discourage optimization beyond the cap. Experiments across multiple datasets show that CapCode detects cheating while preserving performance ranking of models, and CapReward reduces cheating behavior, yielding models that better follow the intended task specification.
Repository-level coding benchmarks such as SWE-bench have driven a rapid surge in the capabilities of coding agents. Yet they usually treat coding tasks as a holistic, binary prediction problem (e.g., resolved or unresolved), neglecting fine-grained agent capabilities such as repository understanding, context retrieval, code localization, and bug diagnosis. In this paper, we introduce SWE-Explore, a benchmark that isolates the evaluation of repository exploration, a critical capability of coding agents. Given a repository and an issue, SWE-Explore asks an explorer to return a ranked list of relevant code regions under a fixed line budget. SWE-Explore covers 848 issues across 10 programming languages and 203 open-source repositories. For each instance, we derive line-level ground truth from independent agent trajectories that successfully solved the same issue, distilling the specific code regions their solution paths actually consulted. We evaluate exploration along coverage, ranking, and context-efficiency dimensions, showing that these metrics strongly track downstream repair behavior. Across a broad set of retrieval methods, general coding agents, and specialized localizers, we find that agentic explorers form a clear tier above classical retrieval. While file-level localization is already strong for modern methods, line-level coverage and efficient ranking remain the key axes differentiating state-of-the-art explorers.
Rishi Desai, Jesse Hu, Joan Cabezas +23cs.SE cs.AI
AI agents are increasingly expected to complete long-horizon workflows that require sustained progress over hours, millions of tokens, and complex environments. Yet current agent benchmarks largely evaluate short-form tasks, such as single pull requests, small tickets, or 5-10 minute exercises, limiting our ability to measure agents' capabilities in planning, long-context understanding, and memory use. We introduce SWE-Marathon, a benchmark of 20 long-horizon tasks spanning software engineering and adjacent technical domains. Each task consists of a unique executable environment, a human-written reference solution, and a multi-layer verification suite. Logged agent attempts average 27.2M total tokens, making SWE-Marathon substantially longer-horizon than existing SWE and command-line agent benchmarks. Current frontier coding agents solve fewer than 30% of tasks. Failures often arise from poor self-verification, self-reported infeasibility, and premature termination. We also observe reward-hacking behavior in 13.8% of rollouts, where agents attempt to exploit the environment or verifier to bypass the intended workflow. SWE-Marathon includes adversarial review of test suites and execution environments, as well as multi-layer checks designed to prevent shortcut solutions. We release SWE-Marathon, evaluation code, and agent trajectories at https://swe-marathon.org/.