Navers lab

Research

Benchmarks, papers, releases, and other lab updates — one place to track how the frontier moves.

  • PPTBench reference reconstruction preview
    Benchmarks

    PPTBench

    A benchmark for faithful, editable reconstruction of scientific flow diagrams: 500 frozen tasks, native PowerPoint artifacts, and staged semantic and visual evaluation.

    Open arXiv Leaderboard GitHub
  • One stage cannot judge a migration: prior benchmarks run a behavioural suite with no migration audit in front of it, while SWE Refactor Bench runs three stages — migration audit, behavioural tests, agentic verification. Below, the four migration categories and the scores of eight frontier models.
    Benchmarks

    SWE Refactor Bench

    Move a whole repository to another stack — language, framework, platform or build system — and change nothing about what it does. A migration starts with its tests already green, so passing them proves nothing: 20 real repositories graded by whether the old stack is gone, whether every recorded behaviour survived, and what six adversarial coding agents can still break.

    Open arXiv Leaderboard Tasks GitHub
  • AI4AI-Bench
    Benchmark

    AI4AI-Bench

    Give a coding agent a real research codebase and four hours to improve how it trains a model. Most tune the settings around the existing method; few rewrite the method itself.

    Open arXiv GitHub Tasks Trajectories
  • BrowserBC method overview
    Papers

    BrowserBC

    BrowserBC distills successful human browser trajectories into reusable skills, helping web agents solve tasks with higher success rates and fewer interactions.

    Open arXiv Code Demo 机器之心
  • OpenChronicle — local-first memory for AI agents
    Open source

    OpenChronicle

    An open, local-first memory layer for LLM agents. OpenChronicle captures structured context from your screen and apps into inspectable Markdown memory — projects, tools, people, decisions — that any tool-calling agent can read, run locally, and share across models.

    Open GitHub 机器之心
  • Frontier-Engineering benchmark composition
    Benchmarks

    Frontier-Engineering

    Open engineering benchmark: agents on optimization across aerospace scheduling, EDA, quantum circuits, fiber optics, battery control, and more. 47 tasks across five engineering categories.

    Open Paper Leaderboard GitHub 机器之心