All pull requests

Pull requests

• Add Hugging Face export for the perturbation benchmark

  #115·dangng2004 opened Nov 2, 2026Contributor·

  ·

• fix: preserve decimal fidelity in quote localization

  #112·ZhijingWu opened Oct 13, 2026·

  ·

• [perturbation] Add experimental PIS evidential-uncertainty evaluator and verifier-reliability analysis (Issue #23) - negative result with diagnosis

  #111·liemtranq opened Oct 11, 2026·

  ·

• Add pilot evaluation for AI review comment quality

  #110·zjrzjrzjr123456-wq opened Sep 26, 2026·

  ·

• Add Anchored Review Format standard for AI review systems

  #106·dangng2004 opened Jul 1, 2026Contributor·1 task·

  ·

• Public-ready cleanup of the quality-proxy and perturbation benchmarks

  #105·dangng2004 opened Jun 28, 2026Contributor·

  ·

• feat: add Requesty as an OpenAI-compatible provider

  #104·Thibaultjaigu opened Jun 25, 2026·

  ·

• Add concrete suggestions to review comments

  #102·Kummu15 opened Jun 17, 2026Collaborator·

  ·

• Add generator selection-bias analysis for the perturbation benchmark

  #100·dangng2004 opened Jun 5, 2026Contributor·

  ·

• Add manual validation samples for perturbations and severity labels

  #99·dangng2004 opened Jun 5, 2026Contributor·

  ·

• Add combined GPT-or-Claude recall of injected errors

  #97·dangng2004 opened Jun 5, 2026Contributor·

  ·

• Overlap (venn) analysis of comments from different models/systems used in paper

  #91·dangng2004 opened May 21, 2026Contributor·6 tasks·

  ·

• Add threshold sweep for perturbation scoring

  #90·dangng2004 opened May 21, 2026Contributor·2 tasks·

  ·

• Add Reviewer 3 (closed-source HTTP API) as a benchmark system

  #85·dangng2004 opened May 15, 2026Contributor·5 tasks done·

  ·

• Real OpenRouter cost in ReviewResult

  #83·dangng2004 opened May 12, 2026Contributor·3 tasks·

  ·

• Add progressive_debate method: adversarial adjudication replaces consolidation

  #37·r-uben opened Mar 11, 2026Contributor·2 of 4 tasks·

  ·
  3