Research.

Original work from the lab. Preprints, accepted papers, and the code that goes with them.

• 2026
  TMLR Accepted

  Anytime Verified Agents: Adaptive Compute Allocation for Reliable LLM Reasoning under Budget Constraints

  A method for letting agents allocate compute adaptively under token, tool-call, and verification budgets, with verifiable reliability guarantees that hold at any cutoff.

  Paper Code
• 2026
  ICML Workshop FAGEN · Under review

  Representational Collapse in Multi-Agent LLM Committees: A Reproducible Failure Mode, Diagnostic, and Partial Fix

  Identifies and measures a representational collapse failure mode in multi-agent LLM committees, and proposes a diagnostic and a diversity-aware partial fix.

  arXiv
• 2026
  arXiv 2603.27467

  TurboAngle: Near-Lossless KV Cache Compression via Uniform Angle Quantization

  A KV-cache compression scheme using uniform angle quantization that retains near-lossless accuracy at large compression ratios, with Triton kernels included.

  arXiv Code