analiza aiqbench
FormulaCode introduces a critical benchmark for evaluating LLM coding agents on real-world repository-scale optimization tasks using multi-objective metrics. The study highlights that current frontier models still struggle to effectively balance correctness with performance improvements across large codebases, marking this as an open challenge in agentic software engineering.
FormulaCode: Evaluating Agentic Optimization on Large Codebases
arXiv:2603.16011v3 Announce Type: replace-cross Abstract: Large language model (LLM) coding agents increasingly operate at the repository level, motivating benchmarks that evaluate their ability to optimize entire codebases under realistic constraints. Existing code benchmarks largely rely on synthetic tasks, binary correctness signals, or single-objective evaluation, limiting their ability to assess holistic optimization behavior. We introduce FormulaCode, a benchmark for evaluating agentic optimization on large, real-world codebases with fine-grained, multi-objective performance metrics. FormulaCode comprises 957 performance bottlenecks mined from scientific Python repositories on GitHub, each paired with expert-authored patches and, on average, 264.6 community-maintained performance workloads per task, enabling the holistic ability of LLM agents to optimize codebases under realistic correctness and performance constraints. Our evaluations reveal that repository-scale, multi-objective optimization remains a major challenge for frontier LLM agents. Project website at: https://formula-code.github.io
Abstract:Large language model (LLM) coding agents increasingly operate at the repository level, motivating benchmarks that evaluate their ability to optimize entire codebases under realistic constraints. Existing code benchmarks largely rely on synthetic tasks, binary correctness signals, or single-objective evaluation, limiting their ability to assess holistic optimization behavior. We introduce FormulaCode, a benchmark for evaluating agentic optimization on large, real-world codebases with fine-grained, multi-objective performance metrics. FormulaCode comprises 957 performance bottlenecks mined from scientific Python repositories on GitHub, each paired with expert-authored patches and, on average, 264.6 community-maintained performance workloads per task, enabling the holistic ability of LLM agents to optimize codebases under realistic correctness and performance constraints. Our evaluations reveal that repository-scale, multi-objective optimization remains a major challenge for frontier LLM agents. Project website at: this https URL
Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
arXiv:2603.16011 [cs.SE]
(or arXiv:2603.16011v3 [cs.SE] for this version)
Text integral sub licența deschisă a publicației · 146 cuvinte
Subiecte conexe
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
[AINews] Muse Spark 1.3 matches GPT-5.6-Sol, confirming Meta Superintelligence as the newest Frontier Lab, >90% discount for training
Latent Space · 3 sept. 2026
[AINews] Gemini 4 Argon: GDM’s answer to Astra/Fable, with 1M output
Latent Space · 1 oct. 2026
AUTOGATE: Automated Clock Gating via Toggling-Aware LLM-based RTL Rewriting
arXiv · 17 iun. 2026