📄 π^4: A Foundation Model for 4D Scene Geometry and Motion
by Yifan Wang (@yyfz321021), Weiguang Zhao et al.
A shared-backbone model that reconstructs geometry, poses, scale, tracks, and motion masks together; motion is predicted in 2D and converted to 3D via reconstructed geometry.
Key points:
• Best overall average rank across geometry benchmarks
• Competitive single- and multiview 2D/3D tracking and motion segmentation
• Uses ~0.5B params, less than half the size of most baselines
Shows that conditioning tracking on a strong frozen geometry model can unify reconstruction and motion without direct 3D motion regression.
1
15
📄 ForceHarness: Turning Force into Action and Physics for Frozen Frontier Models
by Ke Sun, Kevin Yuchen Ma et al.
Closes force control under an API-only model with self-terminating skills, then translates force traces into physical quantities the model can reason over.
Key points:
• Real robot: 90.0% vs 3.3% zero-shot success for GPT-6 Astra
• ForceBench: 87.5% vs 70.0% without force; beats 35.3% best trained
• HiddenPhys: 96.4% vs 62.1% without force; beats 59.3% best trained
Frozen frontier models can handle contact-rich manipulation well when force is exposed as closed-loop skills plus physics-level feedback, not raw sensor values.
1
21
📄 ScaleVLA: Scaling Vision-Language-Action Models to 552 Billion Parameters
by Ruixi Yu, Dairu Liu et al.
Freezes a 552B frontier VLM and adds a trainable action DiT, using one encoder pass per action chunk instead of text reasoning at control time.
Key points:
• LIBERO-PRO: 45.1% vs 31.3% best smaller backbone
• LIBERO-Goal: 87.8% with 10 demos vs 85.6% for 2B with 50
• Real charger insertion: 9/10 vs 4/10 successes over 2B
Frontier-scale pretrained VLM representations can materially improve robot policy data efficiency, robustness, and contact-sensitive control without per-step language reasoning.
1
44
📊 swe-cowork by Tokenwave-AI (@TokenWaveai)
Contains 101 software engineering assignments on real open-source repos with workplace records, chat, and hidden-test evaluation for coding agents.
Key points:
• 101 tasks across 11 primary languages
• 2,712 total requirements; 2,217 held only by colleagues
• 2,701 F2P + 561 P2P hidden test nodes; 221 container images
Useful for benchmarking coding agents on long-horizon repo work with requirement discovery through teammate interaction.
📥 214 downloads
2
36
📄 K2-Horizon: A Frontier Fleet of Open Source Language Models
by K2 Team, Zhengzhong Liu et al.
A fully open LLM fleet plus training data, logs, checkpoints, and code; it pairs MoVA sparse value routing with softplus gating and Uno decoding adapters.
Key points:
• 375B-A23B gets 65.3 on Toolathlon-Verified, beating GPT-5.6 Terra
• 375B-A23B scores 48.4 on SWE-Atlas-QnA, leading open-weight peers
• Uno speeds decoding by up to 2.2x without a separate draft model
Shows sparse value routing can retain dense-model quality at far lower active parameter count while a fully open release enables controlled scaling and sparsity studies.
1
2
80
💻 huawei-csl/MACaQ
Quantizes multimodal models for local inference and exports GGUF or compact checkpoints for llama.cpp, vLLM, Diffusers, and PyTorch.
Key points:
• Quantizes 122B params in 64 min using 32.5 GB device memory
• vLLM W4A4 gets 2.33× prefill and 2.58× decode vs BF16
• MACaQ-R cuts KL on Qwen3.6-35B-A3B from 0.015731 to 0.013367
Shows calibrated low-bit quantization can extend beyond LLMs to a shared workflow for text, image, video, and audio models.
⭐ 337 stars
2
3
512
📊 SO-Arena by SO-Arena
Contains 2,919 text-based evaluation tasks spanning static judgments, tool-use episodes, and web tasks for decision-making models.
Key points:
• 2,919 tasks across 11 domains
• 2,300 static + 282 tools + 337 web tasks
• Mixed-source license; viewer exposes 2,479 public tasks
Useful for benchmarking LLM decision capability and reliability in static, tool, and web settings.
📥 681 downloads
1
21
📄 V-JEPA-LM: Video-Language Models That Read Dynamics Beyond Static, Text-Aligned Representations
by Keuntae Kim, Dongjun Kim et al.
Counterfactual transformed-video supervision makes an LM read motion from frozen V-JEPA features, using one-stage training with far less data and fewer visual tokens.
Key points:
• 4B model: 72.1 avg vs 61.1 best open 4B and 64.8 best open 8B
• At matched budget, counterfactual supervision adds 4.5 avg points and +38 on speed questions
• On held-out speed items, V-JEPA+CF answers 92.8% vs 33.7% with captions+QA
Shows that temporal ability depends on both the encoder's shared dynamics representation and supervision that explicitly requires reading it.
1
47
💻 mit-han-lab/pear
by @songhan_mit
Trains and deploys robot-using reasoning models from simulation only, with async RL, benchmark eval, and a real-robot agent loop.
Key points:
• Targets an open 27B reasoning model trained entirely in simulation
• Supports zero-demonstration training and real-robot deployment
• Covers LIBERO, LIBERO-Plus, LIBERO-Pro, RoboDojo, and RoboLab-120
Shows that large reasoning models can be post-trained in simulation to operate real robots without demonstrations.
⭐ 350 stars
1
1
77
📊 Ropedia-Xperience-2 by ropedia-ai (@ropedia_ai)
Contains 1,500+ metric 3D scene scans plus 4,000+ egocentric recordings with LiDAR, 4-camera video, IMU, body/hand mocap, and text annotations.
Key points:
• 1,500+ scanned scenes from 596 locations
• 4,000+ egocentric recordings across 74 tasks
• 131 hours of first-person video; CC BY-NC 4.0
Useful for training and benchmarking embodied agents on real-world 3D perception, action understanding, and real-to-sim tasks.
📥 6k downloads










