@HuggingPapers

Tweeting interesting papers submitted at https://xitter.cf/t.co/rXX8x0HzXV. Submit your own at https://xitter.cf/t.co/QhbJKXBd4Q, and link models/datasets/demos to it!

 Anywhere
huggingface.co/docs/hub/en/model-release-checklist
Joined March 2025
  • Tweets7.7K
  • Following4
  • Followers22.1K
3.6KMedia
SpatialScout: Can VLM agents actively explore and map 3D scenes? A new benchmark for evaluating vision-language models on active object-level 3D mapping across 80 scenes and 3,140 objects.
1
3
12
1,416
Paper: paperswithcode.co/paper/2610… Dataset: huggingface.co/datasets/jess…

SpatialScout: Can VLM Agents Actively Explore and Map 3D Scenes? | Papers with Code

SpatialScout evaluates VLM agents that actively explore indoor scenes and build object-level 3D maps. Across 80 scenes and 3,140 objects, Claude Opus 5.5 reaches 41.71% mean overlap, but only 3.25%...

paperswithcode.co
1
173
StreamEval: The first benchmark for interactive streaming video generation It evaluates 37 models across 5 tasks with 300 scripted test cases, revealing that reliable request execution remains a central challenge.
3
4
1
21
3,205
Explore StreamEval, the open benchmark for interactive streaming video generation. Evaluate your models, see the leaderboard, and contribute. Paper: paperswithcode.co/paper/2610… Dataset: huggingface.co/datasets/Stre… Leaderboard: streameval.ai/ Code: github.com/StreamEval/Stream…

StreamEval: Evaluating Interactive Streaming Video Generation Models | Papers with Code

StreamEval benchmarks interactive streaming video across five tasks, measuring execution, state consistency, visual fidelity, and audiovisual coordination. Across 37 models, request execution remains...

paperswithcode.co
1
360
DeepSeek just released an upgraded Flash model on Hugging Face Better reasoning, coding, agentic tasks and visuals, with faster execution and shorter outputs at lower cost.
3
29
2,121
1M-token context, MIT license HLE 42.6 · Terminal-Bench 42.3 · DeepSWE 77.0 Model: huggingface.co/deepseek-ai/D…

deepseek-ai/DeepSeek-V4.1-Flash-1017 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co
2
508
AKV: Cache Engineering for Agents A new serving system that lets LLM agents directly manage their KV cache — dropping unused history and repositioning retained keys without rewriting or re-prefilling. Reduces average cost by 27.7% while improving task accuracy by 1.7 points.
6
3
2
41
2,128
Paper: paperswithcode.co/paper/2610… Code: github.com/thu-nics/AKV Project: fuvty.github.io/thinking_yar…

Cache Engineering for Agents | Papers with Code

Cache engineering lets agents control active KV caches without rewriting trajectories or re-prefilling retained context. Rolling drop excludes stale entries and repos adjusts position encodings,...

paperswithcode.co
5
418
Microsoft Research's WildAWM A scalable pipeline that synthesizes executable computer-use agent environments and tasks using only open-source models, yielding 341 long-horizon env-task pairs and Qwen agents that generalize to 8 unseen benchmarks.
5
4
1
43
2,607
Paper: paperswithcode.co/paper/2610… Envs & tasks: huggingface.co/datasets/Chil… Models: huggingface.co/collections/C…

Agent World Model in the Wild: Computer-Use Agent Environment Synthesis | Papers with Code

WildAWM synthesizes executable worlds spanning GUIs, MCP services, and command lines, then filters tasks with rollout, judge-consistency, and counterfactual checks. Its 341 tasks remain difficult for...

paperswithcode.co
4
515
Meta just released MemProBench on Hugging Face 800 frozen memory histories from real-world Ego4D activity, built to study memory-driven proactive assistance: when should an assistant speak up, not just what to say?
2
4
24
1,489
Each history is a byte-preserved caption log with timestamps, episodes, and evidence labels. Dataset: huggingface.co/datasets/face… Code: github.com/facebookresearch/…

facebook/MemProBench · Datasets at Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co
5
511
Sage-A1.5 Shanghai AI Laboratory just released a 35B MoE agent on Hugging Face that reaches trillion-parameter performance: it orchestrates parallel subagents, verifies its own progress, and conducts open scientific discovery.
6
6
42
2,103
Paper: paperswithcode.co/paper/2610… Model: huggingface.co/InternScience… Project: alpha-innovator.github.io/Sa…

Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition |...

Sage-A1.5 composes delegation, parallel subagents, verification, and strategy revision for long-horizon problem solving. Its 35B mixture-of-experts design supports inspectable scientific exploration...

paperswithcode.co
3
533
Paper: paperswithcode.co/paper/2610… Project: ropedia.github.io/S-Agent-2.… Code: github.com/Ropedia/S-Agent-2…

S-Agent 2.0: Spatial Reasoning with a General Coding Agent | Papers with Code

S-Agent 2.0 closes the gap between perception tools and spatial conclusions by combining command-line geometry, persistent scene memory, and plain-language interpretations. It improves Qwen3.8-27B by...

paperswithcode.co
7
443
S-Agent 2.0: Spatial reasoning with a general coding agent A command-line perception toolkit closes the evidence loop — acquire, remember, interpret — boosting a coding agent by +20.5 on ReVSI and +13.9 on MMSI.
4
9
52
2,284
Ropedia W0 A world model that learns loco-manipulation from 10k hours of human egocentric video, then transfers across embodiments, from robot arms to humanoids and from grippers to dexterous hands.
5
11
32
4,932
Paper: paperswithcode.co/paper/2610… Model: huggingface.co/ropedia-ai/Ro… Dataset: huggingface.co/datasets/rope…

Ropedia W0: Scaling Loco-Manipulation World Models across Embodiments with Human Experience |...

Ropedia W0 unifies human and robot egocentric data for cross-embodiment loco-manipulation world modeling. Its kinematic representation combines camera poses with projected skeletons, and 10k hours of...

paperswithcode.co
5
424
Skywork AI's Matrix-Game 4.0 A playable game world rendered by a video model in real time: 480p at 60 FPS on a single RTX 5090, with a key-press-to-display latency of just 112ms.
4
4
2
32
1,888
Paper: paperswithcode.co/paper/2610… Models, weights & game assets (Medieval Castle, Ancient Egypt, City Racing): huggingface.co/Skywork/Matri…

Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds | Papers with Code

Matrix-Game 4.0 separates game execution from visual generation using an executable engine, coding agent, and causal video renderer conditioned on captions and memory. It delivers 480p gameplay at 60...

paperswithcode.co
4
493
Load more