Meta’s GAMUT benchmark measures whether long answers are complete, not just correct
facebook · hf · 2026-07-22
Meta introduces GAMUT, a benchmark for factual completeness in open-ended long-form generation, built around a two-level meta-rubric framework.
Why it matters
- Prior factuality work mostly checks precision: whether claims are correct.
- GAMUT targets the other half: whether an answer includes all the facts it should.
- The framework compiles a structured meta-rubric into a flat checklist of machine-gradable binary rubrics.
Dataset and evaluation
- Contains 1,813 questions grounded in real wearable imagery across 10 domains.
- Each question is paired with an evidence-backed rubric verified by expert annotators.
- Also ships a text-only variant because the framework is modality-agnostic.
Results
- Evaluated 14 frontier and open-weight models.
- The benchmark is described as challenging and discriminative, with the best score at 58.7% from Gemini 3.1 Pro.
Related event: Meta Introduces GAMUT Benchmark for Evaluating Long-Form Text Completeness(2 posts)→
More from Research
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27
- UniAction: unified action conditioning lets one world model handle mixed action types — ZeYanjie · 2026-10-27
- Infini-AI Lab open-sources Hermit, a framework for agent harness evolution and continual learning — BeidiChen · 2026-10-27
- Meta FAIR's Self-Direction turns agent failures into reusable LoRA skills, +13.8pp on Terminal-Bench — karen_ullrich · 2026-10-27