🔴 High Significance
Model Releases
🔴 BitDance: Scaling Autoregressive Generative Models with Binary Tokens — score 75
Sources: huggingface
We present BitDance, a scalable autoregressive (AR) image generator that predicts binary visual tokens instead of codebook indices. With high-entropy binary latents, BitDance lets each token represent up to 2^{256} states, yielding a compact yet highly expressive discrete representation. Sampling fr
Developer Tools
🔴 Experiential Reinforcement Learning — score 95
Sources: huggingface
Reinforcement learning has become the central approach for language models (LMs) to learn from environmental reward or feedback. In practice, the environmental feedback is usually sparse and delayed. Learning from such signals is challenging, as LMs must implicitly infer how observed failures should
🔴 DeepImageSearch: Benchmarking Multimodal Agents for Context-Aware Image Retrieval in Visual Histories — score 85
Sources: huggingface
Existing multimodal retrieval systems excel at semantic matching but implicitly assume that query-image relevance can be measured in isolation. This paradigm overlooks the rich dependencies inherent in realistic visual streams, where information is distributed across temporal sequences rather than c
🟡 Notable
Model Releases
🟡 Nanbeige4.1-3B: A Small General Model that Reasons, Aligns, and Acts — score 65
Sources: huggingface
We present Nanbeige4.1-3B, a unified generalist language model that simultaneously achieves strong agentic behavior, code generation, and general reasoning with only 3B parameters. To the best of our knowledge, it is the first open-source small language model (SLM) to achieve such versatility in a s
🟡 REDSearcher: A Scalable and Cost-Efficient Framework for Long-Horizon Search Agents — score 55
Sources: huggingface
Large language models are transitioning from generalpurpose knowledge engines to realworld problem solvers, yet optimizing them for deep search tasks remains challenging. The central bottleneck lies in the extreme sparsity of highquality search trajectories and reward signals, arising from the diffi
Developer Tools
🟡 STATe-of-Thoughts: Structured Action Templates for Tree-of-Thoughts — score 45
Sources: huggingface
Inference-Time-Compute (ITC) methods like Best-of-N and Tree-of-Thoughts are meant to produce output candidates that are both high-quality and diverse, but their use of high-temperature sampling often fails to achieve meaningful output diversity. Moreover, existing ITC methods offer limited control
Enterprise Adoption
🟡 Accelerating discovery in India through AI-powered science and education — score 50
Sources: lab_blog/DeepMind
Google DeepMind brings National Partnerships for AI initiative to India, scaling AI for science and education
🟢 Incremental
Model Releases
🟢 Query as Anchor: Scenario-Adaptive User Representation via Large Language Model — score 35
Sources: huggingface
Industrial-scale user representation learning requires balancing robust universality with acute task-sensitivity. However, existing paradigms primarily yield static, task-agnostic embeddings that struggle to reconcile the divergent requirements of downstream scenarios within unified vector spaces. F
🟢 Data Darwinism Part I: Unlocking the Value of Scientific Data for Pre-training — score 25
Sources: huggingface
Data quality determines foundation model performance, yet systematic processing frameworks are lacking. We introduce Data Darwinism, a ten-level taxonomy (L0-L9) that conceptualizes data-model co-evolution: advanced models produce superior data for next-generation systems. We validate this on scient
Developer Tools
🟢 InnoEval: On Research Idea Evaluation as a Knowledge-Grounded, Multi-Perspective Reasoning Problem — score 15
Sources: huggingface
The rapid evolution of Large Language Models has catalyzed a surge in scientific idea production, yet this leap has not been accompanied by a matching advance in idea evaluation. The fundamental nature of scientific evaluation needs knowledgeable grounding, collective deliberation, and multi-criteri
🟢 Learning to Configure Agentic AI Systems — score 5
Sources: huggingface
Configuring LLM-based agent systems involves choosing workflows, tools, token budgets, and prompts from a large combinatorial design space, and is typically handled today by fixed large templates or hand-tuned heuristics. This leads to brittle behavior and unnecessary compute, since the same cumbers
📄 New Papers
| Title | Category | Score | Link |
|---|---|---|---|
| Experiential Reinforcement Learning | developer_tool | 80 | Open |
| DeepImageSearch: Benchmarking Multimodal Agents for Context-Aware Image Retrieval in Visual Histories | developer_tool | 62 | Open |
| BitDance: Scaling Autoregressive Generative Models with Binary Tokens | model_release | 56 | Open |
| Nanbeige4.1-3B: A Small General Model that Reasons, Aligns, and Acts | model_release | 38 | Open |
| REDSearcher: A Scalable and Cost-Efficient Framework for Long-Horizon Search Agents | model_release | 29 | Open |
| Enhancing Diversity and Feasibility: Joint Population Synthesis from Multi-source Data Using Generative Models | cs.AI | 0 | Open |
| When Remembering and Planning are Worth it: Navigating under Change | cs.AI | 0 | Open |
| Accelerating Large-Scale Dataset Distillation via Exploration-Exploitation Optimization | cs.AI | 0 | Open |
| Visual Persuasion: What Influences Decisions of Vision-Language Models? | cs.AI | 0 | Open |
| High-Fidelity Network Management for Federated AI-as-a-Service: Cross-Domain Orchestration | cs.AI | 0 | Open |
| Complex-Valued Unitary Representations as Classification Heads for Improved Uncertainty Quantification in Deep Neural Networks | cs.AI | 0 | Open |
| AI-Paging: Lease-Based Execution Anchoring for Network-Exposed AI-as-a-Service | cs.AI | 0 | Open |
| The Information Geometry of Softmax: Probing and Steering | cs.AI | 0 | Open |
| EAA: Automating materials characterization with vision language model agents | cs.AI | 0 | Open |
| Foundation Models for Medical Imaging: Status, Challenges, and Directions | cs.AI | 0 | Open |