SOL: a sample-based distance metric proposed to evaluate diffusion language models
LucaAmb · x · 2026-10-08
Amid ongoing debate about how hard it is to evaluate diffusion language models, Greg Kornhardt's team proposes SOL, a sample-based distance between generated and real text distributions, detailed in a thread. The metric sidesteps per-sample scoring and assesses diffusion LM generation quality at the distribution level.
More from Research
- Commercial detector re-runs NeurIPS AI-text check: 7.3% of 2025 papers flagged vs Pangram's 1% — AltruisticCouple3491 · 2026-10-08
- Burned 300B Tokens with Nothing; Internal Model Broke Through in 3 Hours — burny_tech · 2026-10-08
- NVIDIA's UNREAL paper lets one LLM both retrieve and answer, lifting recall from 49% to 73% — mark_k · 2026-10-08
- NAMVIS: next-scale autoregression beats diffusion for multi-view synthesis, 3x faster — Ramil Khafizov · 2026-10-08
- Neuphonic open-sources NeuDecide: a 43MB audio-to-tool-call model that runs on one CPU thread — TeamNeuphonic · 2026-10-08
- Only 162 of OpenAI's 722 math papers carry Lean-verified proofs — gerardsans · 2026-10-08