AI models and papers of week 41, 2026
2 recorded days folded into one page: the 15 most upvoted papers of the week and the models that held a place on the trending list, from 2 days of model records.
The paper with the most votes this week was “AgentGarten: Code Worlds for Evolving Agents” from MirroS, with 149 upvotes on the list of October 9, 2026. 12 of the 15 papers recorded came with a code repository.
Papers of the week
- 149
AgentGarten: Code Worlds for Evolving Agents
Fri Oct 9 · MirroSarXiv 2610.12374 (opens in a new tab)Code (opens in a new tab)
- 135
Learn2Play Bench: How Well Do LLM Agents Learn from Experience in Unfamiliar Environments?
Fri Oct 9 · National University of SingaporearXiv 2610.08215 (opens in a new tab)Code (opens in a new tab)
- 129
TokenRouter: Efficient Serving System for Token-Level LLM Routing
Fri Oct 9 · Tsinghua-NICS-EFCarXiv 2610.12242 (opens in a new tab)Code (opens in a new tab)
- 105
From Traces to Agentic Worlds: Agentic Language World Models for Interactive Environment Simulation
Fri Oct 9 · Nanyang Technological University SingaporearXiv 2610.06100 (opens in a new tab)Code (opens in a new tab)
- 71
SuperNav: An Agentic Navigation System for Any Task in Any Scene
Fri Oct 9 · zju3dvarXiv 2610.12126 (opens in a new tab)Code (opens in a new tab)
- 71
MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement
Fri Oct 9 · Xiaomi MiMoarXiv 2610.11959 (opens in a new tab)
- 52
Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction
Fri Oct 9 · KAIST AIarXiv 2610.12299 (opens in a new tab)Code (opens in a new tab)
- 47
In-context Robot Learning Made Simple: A Democratized Recipe for Manipulation Tasks
Fri Oct 9arXiv 2609.38173 (opens in a new tab)Code (opens in a new tab)
- 39
OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D Cinemagraphs
Fri Oct 9 · Alaya LabarXiv 2610.12461 (opens in a new tab)Code (opens in a new tab)
- 38
Beyond Spatio-Temporal Priors: A Generalizable Approach for Dense Correspondence Matching
Fri Oct 9 · The University of Hong KongarXiv 2610.12421 (opens in a new tab)Code (opens in a new tab)
- 38
U-Space: Uncovering When and Why Uncertainty Arises in Language Models
Fri Oct 9 · Multimodal AI Lab - Technical University of DarmstadtarXiv 2610.09087 (opens in a new tab)Code (opens in a new tab)
- 37
DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training
Fri Oct 9arXiv 2610.12468 (opens in a new tab)Code (opens in a new tab)
- 37
MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers
Fri Oct 9 · Tencent HunyuanarXiv 2610.06801 (opens in a new tab)Code (opens in a new tab)
- 34
TestPrism: Rethinking Test Evaluation Beyond a Single Reference
Fri Oct 9 · NJU-LINK LabarXiv 2610.12289 (opens in a new tab)
- 34
Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks
Fri Oct 9 · University College LondonarXiv 2610.11794 (opens in a new tab)
Models of the week
2 of 2 days · best No. 1 · Feature extraction · 45.6K downloads / 30 d
Cloudflare/clef (opens in a new tab)
2 of 2 days · best No. 2 · Image text to text · 13.6K downloads / 30 d
2 of 2 days · best No. 2 · Text generation · 33.7K downloads / 30 d
abenzerps/Qwen-Image-2.1-Uncensored-GGUF (opens in a new tab)
2 of 2 days · best No. 4 · Text to image · 2.1M downloads / 30 d
Aleph-Alpha/Kolibri-1 (opens in a new tab)
2 of 2 days · best No. 5 · Text generation · 10.5K downloads / 30 d
Venastine-Research/Xing4.0-29B-A4B-GGUF (opens in a new tab)
2 of 2 days · best No. 6 · Text generation · 45.2K downloads / 30 d
Qwen/Qwen-Image-2.1-Turbo (opens in a new tab)
2 of 2 days · best No. 6 · Text to image · 2,138 downloads / 30 d
2 of 2 days · best No. 7 · Image to video · 1.72M downloads / 30 d
2 of 2 days · best No. 9 · Image text to text · 6.77M downloads / 30 d
Cloudflare/clef-flash (opens in a new tab)
2 of 2 days · best No. 10 · Image text to text · 20.7K downloads / 30 d
canberkkkkkk/ema-lightning (opens in a new tab)
2 of 2 days · best No. 10 · Text to speech · 15.7K downloads / 30 d
convaiinnovations/laya (opens in a new tab)
2 of 2 days · best No. 12 · Text classification · 42.9K downloads / 30 d
Cactus-Compute/whistle (opens in a new tab)
2 of 2 days · best No. 13 · Automatic speech recognition · 12.8K downloads / 30 d
ConwayResearch/Underdog-Saluki-27B-1.0 (opens in a new tab)
2 of 2 days · best No. 13 · Text generation · 59.2K downloads / 30 d
LiquidAI/d1-3B (opens in a new tab)
2 of 2 days · best No. 14 · Image text to text · 9,088 downloads / 30 d
About the archive
How far back does the archive go?
Model lists exist from the first day the collector ran: Hugging Face does not publish past trending lists, so they cannot be rebuilt. Paper lists for earlier days were rebuilt from the Hugging Face daily papers archive, and those days are marked "papers only".
What is kept for each day?
The leading models with their task, likes and downloads on that day, the models held back, and the leading papers with their upvotes. Abstracts are not duplicated in the archive; each paper links to arXiv.
Do archived numbers change?
No. A day’s record is the last collection of that day and is not rewritten afterwards. Likes and downloads you see on an archived day are the values on that day, not today’s.
How are weeks defined?
ISO 8601 weeks: Monday to Sunday, numbered so that week 1 contains the first Thursday of the year. The address /week/2026-40 is week 40 of 2026.