Broken on arrival: many quantized LLM artifacts in public registries don't work
Independent-Researcher · hf · 2026-10-31
A study of quantized small LLM files in public model registries finds many artifacts are non-functional on arrival due to uncalibrated quantization recipes. A 15-task pass-rate heatmap across models (codellama, deepseek-coder, granite-code, llama, mistral, qwen2.5-coder, starcoder2, etc.) flags suspect files, some passing zero tasks. The authors recommend screening plus calibration before release.
More from Research
- Professor runs Pangram on 50 pre-ChatGPT self-published novels: zero false positives — TuhinChakr · 2026-10-31
- Researchers push back on Samuelson brief with Pangram audit: 0.04% false positives — TuhinChakr · 2026-10-31
- Stanford's TaskNet builds occupational task inventories from worker interviews at scale — stanfordnlp · 2026-10-31
- Aligning image and text embeddings without paired data: Wasserstein Procrustes and rotation papers — phillip_isola · 2026-10-31
- Phillip Isola demos aligning image and text embeddings with just a single rotation — phillip_isola · 2026-10-31
- Arena Offers Up to $50,000 Per Project for Academic Research on AI Evaluation — arena · 2026-10-31