GitHub 新星
10 条
arXiv 论文
10 条
-
1
Text-to-Image Generative Modeling Without Language Pretraining
Modern text-to-image (T2I) systems use internet-scale language pretraining for text-conditioned image generation. We ask whether language pretraining is necessary for strong T2I generation. We introduKeshigeyan Chandrasegaran, Yue Zhao, Armin W. Thomas
-
2
OpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal Memory
Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through internal reasoning alone: our controlled studYu Zhou, Cheng-Fu Yang, Rui Sun
-
3
Reinforcement Learning with Visually Attributable On-Policy Self-Distillation
Multimodal reinforcement learning with verifiable rewards (RLVR) provides reliable task-level supervision, yet offers limited guidance for localizing errors in visual reasoning. On-policy self-distillShiao Xie, Siyu Chen, Jianwei Lv
-
4
CoopUQ: Robust Multi-Robot Social Navigation via Uncertainty-Aware Cooperative Forecasting
Reinforcement learning enables mobile robots to make fast and scalable navigation decisions in constrained human-shared environments. However, directly extending single-robot approaches to multi-robotZehao Wang, Jianpeng Yao, Yuping Wang
-
5
Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy
In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-eval to quantify physical RSI through a fZekai Li, Jiayi Yu, Jinlu Zhang
-
6
WeAct: Dynamic Weight Switching as Agent Actions
Long-horizon agents must learn heterogeneous capabilities from interleaved trajectories. Updating shared parameters to learn new capabilities can degrade previously acquired ones, resulting in catastrRui Shi, Xunjian Yin, Jiacheng Sang
-
7
Train An Imaginative Agent
A seemingly reasonable push in Sokoban can make the puzzle unsolvable. To avoid such mistakes, an agent has to anticipate the consequences of its actions before execution, especially when feedback froJiaxin Ge, Junyi Zhang, Tianning Chai
-
8
AnyGoal: On World Model Generalization Beyond Expert Rollouts
The goal of a world model is to capture the dynamics of an environment. This has the great appeal that it allows agents to plan and simulate a course of actions with the help of a world model. For theHeejeong Nam, Divyat Mahajan, Quentin Le Lidec
-
9
LatComm: Reusable Context Compression for Latent Communication between Large Language Models
Systems that combine several large language models (LLMs) often use one context for many questions. When the context is passed between models as text, the receiving model processes all of its tokens aHuawei Lin, Jinting Liu, Tan Yu
-
10
Newton-Based Methods for Bilevel Optimization Beyond Lower-Level Strong Convexity
Bilevel optimization provides a framework for machine learning tasks such as hyperparameter optimization, data reweighting, and meta-learning. Existing Newton-based bilevel methods typically rely on tYouran Dong, Junfeng Yang, Wei Yao
Hacker News
10 条
- 1 DeepInfer: High-performance inference framework for DeepSeek models
- 2 EMudhra Intermediate CA Compromise
- 3 SimCity, but with Permits
- 4 Voxtral 3
- 5 Apple broke an essential presentation feature of iPads
- 6 Facebook Pages require paid subscription to post more than two links per month
- 7 Tsrs: A Port of TypeScript to Rust
- 8 Google Cloud outage affects almost all banking and other operations in Europe
- 9 Ofcom concedes defeat in 4chan case
- 10 How I got over 300 Tor exit nodes removed