Skip to content
Papers.
Updated 12h ago

AI and data research papers, by lab

40 notable papers from 15 labs and companies, 10 of them from the last 12 months, with 602k citations between them. One plain line each.

Papers by lab, 2017 to nowDot size is citations. Hover a dot, or focus the chart and use the arrow keys.
2017'172018'182019'192020'202021'212022'222023'232024'242025'252026'26

Latest papers

Most cited

40 of 40 papers, newest first

PaperCitations
RRSI: Regularized Recursive Self-Improvement of Agent HarnessesRRSI keeps self-improving agent harnesses from memorising their training tasks, so gains carry over to new benchmarks.Google · Sep 20260
DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache CompressionA 552B mixture-of-experts model with a 1M-token context, built to shrink the KV cache for agent workloads.DeepSeek · Sep 20267
StudentSim: Training LLM-based Student SimulatorsStudentSim trains per-student simulators that answer like a given learner and change their answers under a tutor's guidance.Microsoft · Sep 20260
On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training StabilityQwen3.8-Flash-Next: a 125B mixture-of-experts model with 6B active that nearly matches its 397B predecessor at 1/9 the training compute.Alibaba (Qwen) · Aug 20266
Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and RecipesControlled experiments on when text, image understanding and image generation help or compete when trained together.Meta · Aug 20262
Gemma 4 Technical ReportOpen multimodal models from 2.3B to 31B parameters, with a thinking mode and image and audio input.Google DeepMind · Jul 2026106
Agents' Last ExamOver 1,000 real, checkable professional tasks written with 250+ industry experts; the hardest tier is far from solved.UC Berkeley · Jun 202615
Cosmos 3: Omnimodal World Models for Physical AIOne model that reads and generates text, images, video, audio and robot actions for physical AI.NVIDIA · Jun 202690
MolmoAct2: Action Reasoning Models for Real-world DeploymentA fully open robot action model, released with new datasets including 720 hours of two-arm teleoperation.Ai2 · May 202644
Strategic Navigation or Stochastic Search? How Agents and Humans Reason Over Document CollectionsOn 2,250 questions over 800 PDFs, the best agents match human accuracy but lean on brute-force search rather than planning.Snowflake · Mar 20261
Why Language Models HallucinateModels hallucinate because training and benchmarks reward confident guessing over saying they do not know.OpenAI · Sep 2025329
Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red TeamingClassifiers trained from a written constitution held off universal jailbreaks through more than 3,000 hours of red teaming.Anthropic · Jan 2025199
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement LearningReinforcement learning taught a model long step-by-step reasoning on par with OpenAI o1.DeepSeek · Jan 20255,691
Alignment faking in large language modelsClaude 3 Opus sometimes went along with a training goal it disagreed with, to avoid being changed, without being told to.Anthropic · Dec 2024342
SwiftKV: Fast Prefill-Optimized Inference with Knowledge-Preserving Model TransformationSwiftKV lets prompt tokens skip later layers and merges KV caches, cutting prefill cost for long-prompt workloads.Snowflake · Oct 202418

Questions

What is Papers?

A calm list of notable data and AI research papers from the labs the fru.dev sites follow: OpenAI, Anthropic, Google, Google DeepMind, Meta, Microsoft, NVIDIA, Databricks, Snowflake, DeepSeek, Alibaba (Qwen), Ai2 and the universities behind the field's core methods. Each paper gets its lab, date, arXiv link, topic and one plain line on what it shows.

Where do the citation counts come from?

From the free Semantic Scholar Graph API, read once a week. "Influential citations" are Semantic Scholar's count of citing papers that build on the work rather than mention it. Upvotes are from Hugging Face papers. Every change to a count is kept on the Changes page.

How are new papers found?

Every Monday the refresh reads the last week of Hugging Face daily papers and keeps papers whose organisation is a tracked lab and that other readers upvoted. New finds appear with an Unverified badge until checked by hand.

Who writes the one-line summaries?

The seed papers were summarised by hand from their abstracts. New finds start with Hugging Face's own summary, marked as such, until they are rewritten in plain words. The line is a pointer; the paper is the source.

What are the most cited AI papers here?

Attention Is All You Need (the Transformer), BERT, GPT-3, InstructGPT and LoRA lead on citations. The papers page sorts by citations, date or upvotes.

Can I use the data in my own tools or agents?

Yes. There is a public JSON API with an OpenAPI spec, llms.txt for language models, and a For AI agents page with copy-paste examples. Please cite Papers (papers.fru.dev).

Something is wrong or missing. How do I fix it?

Every paper and lab page has Suggest a correction, and the listing pages have Suggest a missing paper. Each suggestion is checked against the source before anything changes.

Sources: each lab's own papers and arXiv, with citation and upvote counts from Semantic Scholar and Hugging Face. One-line summaries are for orientation, not a substitute for the paper. Logos via logo.dev; trademarks belong to their owners.

New papers by email

Monday afternoons, only in weeks with new papers from the labs.

Double opt-in. Unsubscribe any time.