GPT-5.2 solves a COLT 2022 open problem the researcher had chased since 2016
kfountou · x · 2026-09-25
Kimon Fountoulakis reports that a paper co-written with GPT was accepted to NeurIPS, solving the open problem he had carried since 2016 and published as an open question at COLT 2022: bounding the total work of accelerated methods for L1-regularized PageRank and showing classical accelerated methods can beat non-accelerated ones in a non-trivial sense.
Prior solutions existed but none analyzed classical accelerated gradient methods—using one gradient per iteration—which are the most practical yet notoriously hard to analyze here. He says he's glad GPT took that burden off his shoulders.
More from Models
- Relace Is Now the Cheapest DeepSeek v4.1 Flash Provider on OpenRouter — ilyasu · 2026-09-25
- Jev matches year-old top models on global geographic understanding, maps extracted — zetalyrae · 2026-09-25
- Claude Opus 5.5 tops SimpleBench with 88.4% score — Profanion · 2026-09-25
- Anthropic resumes billing for safety-blocked requests; 99.7% of users unaffected — ClaudeDevs · 2026-09-25
- Developer claims: nothing holds back Claude models like Claude Code itself — tokenbender · 2026-09-25
- Blogger: Opus 5.5's strength suggests xAI's rumored Astra is smaller than believed — scaling01 · 2026-09-25