cập nhật 55 phút trước

⏱ Độ trễ thật của tab này: trung bình 6.8 giờ sau khi CVE được công bố (n = 5263, sớm nhất 0.1 giờ). Số âm nghĩa là ta thấy trước ngày công bố.

20
Multi-Token Prediction Done Right
📄 Paper HF Daily Papers 23 ngày trước

Multi-Token Prediction (MTP) is now a common ingredient in language-model pre-training, as it improves downstream performance and enables faster inference through speculative decoding. We identify a flaw in standard MTP training, in which a residual component of the MTP gradient …

🅲 hạng C
20
No Channels Left Behind: Understanding Diffusion Acceleration through Residual Gates
📄 Paper HF Daily Papers 21 ngày trước

Training acceleration methods for diffusion transformers, such as REPA and dispersive regularization, are commonly explained by additional supervision, spatial-structure priors, or improved feature geometry. We revisit this view and uncover a striking internal change: the AdaLN-Z…

🅲 hạng C
20
CrossSWE: Benchmarking Cross-Repository Joint Development for Coding Agents
📄 Paper HF Daily Papers 19 ngày trước

As coding agents achieve increasingly high success rates on software engineering benchmarks, recent efforts have sought to increase task complexity through longer horizons and more extensive code modifications. However, real-world software development often spans multiple interde…

🅲 hạng C
20
EvoWorld: Towards Out-of-Sight World Evolution in Generative Video Models
📄 Paper HF Daily Papers 15 ngày trước

Generative video models are increasingly used as visual world simulators, yet their primary strength lies in rendering plausible pixels rather than reasoning about how a world should evolve. Coherent world evolution requires more than visual fidelity: when an event occurs, a mode…

🅲 hạng C
20
Longtail-SWE: High-Quality SWE Tasks for Underrepresented Languages
📄 Paper HF Daily Papers 13 ngày trước

Challenging repository-level benchmarks remain limited for many programming language ecosystems, where filtering scarce candidates with ambiguous descriptions or defective tests can discard valuable engineering problems. We introduce Longtail-SWE, a benchmark of 202 issue-resolut…

🅲 hạng C
20
DySENSE: Dynamic Spatiotemporal ENergy Synthesis and Emulation with Satellite Imagery
📄 Paper HF Daily Papers 7 ngày trước

Cities account for over two-thirds of the world's energy and more than 70% of global carbon dioxide emissions. City energy is essential for achieving the 2050 net-zero target and the UN's Sustainable Development Goals 7 and 11. Although existing studies have made progress, they a…

🅲 hạng C
20
GlyphAtlas: Disentangling Stroke Detail from Layout for Scene Text Synthesis
📄 Paper HF Daily Papers 5 ngày trước

Scene text synthesis requires stroke-accurate characters rendered in a specified layout, as well as visual coherence with the surrounding scene. To achieve this, existing methods commonly pre-render the target text at its intended locations to form a glyph image that jointly spec…

🅲 hạng C
20
PanDWM: Panoramic 4D Driving World Models
📄 Paper HF Daily Papers 5 ngày trước

Driving world models usually generate videos from six pinhole cameras with separate optical centres and small overlaps, and must learn to keep these views consistent. We present PanDWM, which stitches them into one 360^circ panoramic band. Camera boundaries then fall inside one i…

🅲 hạng C
20
Every Weight Is a Path of Shared Discrete Decisions
📄 Paper HF Daily Papers 29 ngày trước

We introduce TWIG (Tree-structured Weights with Inherited Gates), a representation in which every weight of a neural network is the end of a path of small discrete decisions: a sign, a shift by a power of two, a zero gate, an index into a shared table. Weights that belong togethe…

🅲 hạng C
20
Qwen-Image-PE: Learning Visual Design for Text-to-Image Prompt Enhancement
📄 Paper HF Daily Papers 27 ngày trước

Learning visual design is essential for prompt enhancement (PE) to turn underspecified user requests into coherent visual specifications for text-to-image (T2I) generation. Existing methods preserve user intent and improve overall visual appeal, with less explicit emphasis on fin…

🅲 hạng C
20
Principia: Test-Time Hypotheses to Principle Construction for Scientific Discovery
📄 Paper HF Daily Papers 28 ngày trước

Scientific discovery requires identifying the scientific principles underlying observed phenomena to guide further investigation. As these empirical principles are often unknown in complex discovery processes, they are constructed via a hypotheses-to-principle pipeline to obtain …

🅲 hạng C
20
Probing Spatial State Representations in Visual Foundation Models
📄 Paper HF Daily Papers 21 ngày trước

An observer sees only part of the surrounding world, yet must keep track of what lies beyond the current view. Spatial perception therefore involves more than recognizing individual images: it calls for a persistent spatial state that is updated under self-motion and used to guid…

🅲 hạng C
20
FLAMIA: FLAsh MIcroscaling Attention for Fast and Causally Safe Training on Blackwell
📄 Paper HF Daily Papers 17 ngày trước

Attention remains a substantial computational cost in Transformer training, particularly at long context lengths. NVIDIA Blackwell GPUs provide high-throughput low-precision matrix multiplication, but exploiting it in attention requires addressing quantization overhead, causal in…

🅲 hạng C
20
Recursa: Toward Recursively Self-Improving Foundation Model Systems
📄 Paper HF Daily Papers 18 ngày trước cũng có trên: arXiv cs.AI

Foundation models in changing environments can encounter unfamiliar domain knowledge that their original training did not cover. However, new domains rarely provide large amounts of expert-created training data or reliable supervisory signals for further training. We introduce Re…

🅲 hạng C
20
VideoMemBench: How Well Do MLLMs Remember in Video Conversations?
📄 Paper HF Daily Papers 17 ngày trước

Multimodal large language models (MLLMs) can now process hour-long videos, making extended multi-turn dialogue over a single video a natural mode of use. Sustaining such a dialogue requires more than answering each question in isolation: the model must retrieve and reuse informat…

🅲 hạng C
20
D3D: Decoupled Diffusion Drafting for Asynchronous Speculative Decoding
📄 Paper HF Daily Papers 16 ngày trước

Target-conditioned diffusion drafters predict candidate tokens in parallel, yet each new block must wait for updated target-model hidden states. This forces drafting and verification to run sequentially, limiting speedup even though each block is generated in one pass. We introdu…

🅲 hạng C
20
EDGAR: Efficient Diffusion-Driven Gaussian Avatar Refinement via Synthetic Failure
📄 Paper HF Daily Papers 15 ngày trước

Recent advances in Gaussian splatting have enabled efficient and photorealistic head avatar reconstruction, yet monocular video-based or image-based methods still suffer from under-constrained geometry and texture, leading to artifacts such as blurring, stretching, and temporal i…

🅲 hạng C
20
WorldDraft: If Code Builds the World, Can Video Models Just Render It?
📄 Paper HF Daily Papers 8 ngày trước

Multimodal coding agents decouple 4D world construction from visual generation, repurposing video models from world generators toward generative renderers. This division of labor raises a fundamental question: if code builds the world, can video models just render it? A reliable …

🅲 hạng C
20
Unbundling Multimodal Vision Encoding
📄 Paper HF Daily Papers 8 ngày trước

How multimodal language models should encode visual input remains up for debate. Most models use a pretrained vision encoder, while some recent recipes train the encoder from scratch with the language model or remove it entirely. Vision encoders bundle together three design choic…

🅲 hạng C
20
MEM-SQL: Memory-Empowered Text-to-SQL
📄 Paper HF Daily Papers 7 ngày trước

Text-to-SQL systems translate natural language questions into SQL queries over increasingly complex databases. However, widely used benchmarks evaluate questions independently, overlooking settings where earlier queries over the same database can supply joins or calculations for …

🅲 hạng C
20
Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?
📄 Paper HF Daily Papers 23 ngày trước

Financial agents must form forecasts and investment judgments from decision-time evidence, with predictions evaluated against subsequent outcomes. Existing benchmarks assess financial knowledge, research workflows, or forecasting, but evaluating the full research process requires…

🅲 hạng C
20
Solv3R: 3D Looped Transformer as a Multi-view Solver
📄 Paper HF Daily Papers 17 ngày trước

Classical structure-from-motion reconstruction proceeds iteratively, refining camera poses and scene geometry from a current estimate. This allows maps to easily expand without recomputing already computed views, or geometry estimates to improve as more observations are provided.…

🅲 hạng C
20
On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer
📄 Paper HF Daily Papers 15 ngày trước

Prefill-to-Autoregressive Cross-model Transfer (PACT), a form of latent communication between language models, combines a larger model's prompt representations with a smaller model's efficient generation, offering a promising way to balance capability and inference cost. However,…

🅲 hạng C
20
Factored One-sided Adam-Moment for Practical and Scalable SOAP
📄 Paper HF Daily Papers 8 ngày trước

Second-order optimizers (e.g., Shampoo, SOAP) have emerged as compelling alternatives to Adam/AdamW and promise faster per-step convergence than AdamW in transformer training. However, they still remain costly in practice, and even SOAP still pays substantial optimizer-state memo…

🅲 hạng C
20
JVT: Just Video Transformer makes pixel-space video diffusion simple
📄 Paper HF Daily Papers 7 ngày trước

Direct diffusion modeling of high-dimensional raw video data is challenging, and existing pixel-space methods often rely on specialized architectures. We introduce JVT (Just Video Transformer), a simple large-patch DiT-style Transformer for pixel-space video diffusion, without a …

🅲 hạng C
20
Rethinking Sequence Compression with Test-time Regression
📄 Paper HF Daily Papers 28 ngày trước

Sequence compression is a fundamental operation and a critical information bottleneck in efficient language and multimodal models, and yet prevailing approaches based on pooling or learned projections remain underperforming. We revisit sequence compression through the lens of tes…

🅲 hạng C
20
On-Policy Self-Training: Data-Free Self-Improvement in Large Language Models
📄 Paper HF Daily Papers 28 ngày trước

On-policy self-distillation enables post-training foundation models without relying on stronger external teachers. However, existing approaches rely on ground-truth question-answer pairs, human-curated problem sets, or other supervised signals. In this work, we reformulate on-pol…

🅲 hạng C
20
ReGE: Stronger Agents through Experience Synthesis
📄 Paper HF Daily Papers 27 ngày trước

Repeated attempts at interactive tasks uncover facts, useful operations, and failures that can inform subsequent execution. We introduce ReGE, a framework for turning this recorded experience into guidance for a fresh attempt. Its main implementation, ReGE-Syn, uses a single coac…

🅲 hạng C
20
Beyond the Estimate: Wiener-Informed Decoding for Easier Training of Pixel-Space Flow Models
📄 Paper HF Daily Papers 25 ngày trước

Learning velocity fields directly in pixel space is difficult and slow. We show that coarse second-order image statistics can ease this task: frozen Wiener estimates improve velocity learning, with gains extending beyond the linear predictions they explicitly provide. Moreover, w…

🅲 hạng C
20
IntReal: A Real-World Dataset and Benchmark for Multimodal Interleaved Generation
📄 Paper HF Daily Papers 23 ngày trước

Humans routinely consume multimodal content, where textual and visual information is naturally interleaved. Although Unified Multimodal Models (UMMs) can jointly generate text and images, multimodal interleaved generation remains largely underexplored. Progress in this area is hi…

🅲 hạng C