WatchAIChecks hourly
Model launches, research, partnerships and funding from OpenAI, Google DeepMind, Anthropic, Meta AI, Mistral and NVIDIA.
82 subscribers
DiscoverSee all in AI →
Frontier AI Research Weekly
The week’s most important papers on large language models: reasoning, agents, evaluation and alignment — with real results, not position pieces.
82Subscribers
1Results so far
Oct 2026Created
1d agoLast update
About
The week’s most important papers on large language models: reasoning, agents, evaluation and alignment — with real results, not position pieces.
Overview
- Every Monday at 09:00 (UTC), one issue with up to 12 of the best finds since the last one.
- Each find comes with a short summary and why it matched; the issue opens with a few lines on what stood out.
- In your feed, by email, and as a push notice if you turn it on.
Topics
Large language modelsReasoningAgentsEvaluation and benchmarksAlignment and safetyTraining methods
Exclusions
- ✕ Surveys and position papers
- ✕ No experiments or results
- ⚑ No code or weights released — kept, but marked
Sources
arXiv · Computation and LanguagearXiv · Machine LearningarXiv · Artificial Intelligenceand other sources
Listed sources are searched first, then the open web.1issue so far
82subscribers
1d agolast update
DigestEvery Monday, 09:00 UTC
Exact rules
Every Monday at 09:00, send me an issue of up to 12 items with Large language models, plus at least one of Reasoning, Agents, Evaluation and benchmarks, Alignment and safety or Training methods. Skip surveys and position papers or no experiments or results. Keep but flag no code or weights released. Search the open web, starting with cs.CL, cs.LG and cs.AI. Matching is strict. Anyone can find and subscribe to it in Discover.
DeliversDigest · Every Monday, 09:00 UTC
TracksThe week’s most important papers on large language models: reasoning, agents, evaluation and alignment — with real results, not position pieces.
Must matchLarge language models
At least one ofReasoning, Agents, Evaluation and benchmarks, Alignment and safety, Training methods
Strictnessstrict — only results that meet every rule
SkipsSurveys and position papers; No experiments or results
Keeps with a warningNo code or weights released
ReadsAutomatic, plus cs.CL (preferred), cs.LG (preferred), cs.AI (preferred)
More in AI
Streams that already watch this field. Follow one as it is — it costs nothing extra.
RadarAIDaily batch, 10:00 UTC
New open-source frameworks and tools for building LLM agents, as they appear on GitHub and Hugging Face.
79 subscribers
RadarAIAs found
New open-weights language and multimodal models from the major labs, as they land on Hugging Face — with size, license and reported results.
79 subscribers
Can’t find what you need? Describe it in a sentence and Monstream builds the stream for you.Create your own →