Stanford paper measures LLM articulation upper bounds, prompt optimizer gains up to 14%
stanfordnlp · x · 2026-10-17
Researchers at Stanford NLP (Alexander Spangher, Daniel E. Ho, Diyi Yang, Noah Goodman, Sanmi Koyejo) published "Articulation Upper Bounds for Language Models," an attempt to quantify how much meaning language can carry between LLMs.
- Optimal messages: a prompt optimization algorithm that beats existing methods by up to 14% on benchmarks while accurately estimating prompt upper bounds.
- Transmission measure: a sender LLM labels text, a receiver tries to recover the sender's prompt, and recovery accuracy correlates strongly with human label error.
- Using observational scaling laws, they study how a concept's transmissibility grows as model capabilities scale.
More from Research
- AuroraASR-1M: A 1MB Speech Recognition Model Running 436x Real-Time on One CPU Thread — Tall_Abrocoma_3533 · 2026-10-18
- Lattice Gaussian distributions: closed-form KL divergence via Riemann theta functions — FrnkNlsn · 2026-10-18
- Megabake compiler hits 5.5x torch.compile speedup on SmolLM — ricklamers · 2026-10-18
- DeepMind's AlphaProtein Novo designs enzymes for chemistry nature never evolved — anthara_ai · 2026-10-18
- mlabonne's llm-course passes 83k GitHub stars with roadmaps and Colab notebooks — mlabonne · 2026-10-18
- Kaggle LLM attribution contest: GLM text is flagged as Claude 44% of the time — HarveenChadha · 2026-10-18