PulseAugur
EN
LIVE 07:48:45

ViBench 2.0 benchmark evaluates AI agents on long-horizon software building

By PulseAugur Editorial · [1 sources] ·

Researchers have introduced ViBench 2.0, a new benchmark designed to evaluate long-horizon software building capabilities for AI agents. Unlike existing benchmarks that focus on single, fully specified tasks, ViBench 2.0 assesses the development of entire products through successive feature requests, mirroring real-world user interaction. Initial evaluations of nine models revealed a significant gap between passing test plans and delivering a fully functional product, with no open-weight models achieving full product delivery. However, progress is rapid, with one model family showing substantial improvements in success rates and cost reduction over three releases. AI

IMPACT Establishes a new standard for evaluating AI agents in complex, long-term software development tasks.

RANK_REASON The cluster describes a new benchmark for evaluating AI agents on software building tasks, presented in an academic paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

ViBench 2.0 benchmark evaluates AI agents on long-horizon software building

How we ranked this

Signal score
27 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The cluster describes a new benchmark for evaluating AI agents on software building tasks, presented in an academic paper. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Daniel Furman, Preeya Kirani, Ibrahim Sheikh, Zhen Li, Peter Zhong, Michele Catasta ·

    ViBench 2.0: A Benchmark for Long-Horizon Software Building

    arXiv:2611.05620v1 Announce Type: new Abstract: On Replit, our agentic engineering platform, users now build entire products without writing code, through successive feature requests to an agent. Yet existing coding benchmarks evaluate single, fully specified tasks, and many are …

RELATED ENTITIES

RELATED TOPICS