Hugging Face Daily Papers
500 articles archived · Visit source ↗ · RSS
-
Hugging Face Daily Papers research 1mo ago
Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding
Abstract LatentStream introduces a progressive latent working memory framework that internalizes streaming visual evidence into compact evolving tokens for continuous reasoning. Generated by thinkingmachines/Inkling-Small Streaming video understanding requires multimodal large…
27 -
The Missing Temporal Link: Temporal Context Routing for Script-Driven Audio-Video Generation
Abstract Temporal Context Routing improves script-aligned timing of shots and dialogue in joint audio-video generation by mapping structured script timing onto shared video-audio temporal axes. Generated by thinkingmachines/Inkling-Small Joint audio-video generation models have…
20 -
FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlow
Abstract FlashRender accelerates generative video rendering via representation alignment, a mean-flow objective, and on-policy distillation to achieve high-quality few-step camera-controlled synthesis. Generated by thinkingmachines/Inkling-Small We present FlashRender, a…
18 -
WorldReward: Reward Modeling for Camera-Conditioned World Models
Abstract WorldReward is a vision-language reward model that evaluates camera-conditioned world models by aligning video chunks with actions and aggregating preferences for both execution consistency and visual quality. Generated by thinkingmachines/Inkling-Small…
36 -
Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
Abstract Random eviction of reasoning tokens matches selective KV cache compression because reasoning traces are self-protecting through redundancy, making scoring unnecessary once prompts are preserved. Generated by thinkingmachines/Inkling-Small Large language models achieve…
22 -
Rethinking On-Policy Distillation of Large Language Models II: One Training Example
Abstract On-policy distillation improves over hundreds of steps from a single query by rapidly covering teacher states, yet student alignment remains slow, indicating the method is algorithm-starved rather than data-starved. Generated by thinkingmachines/Inkling-Small On-policy…
18 -
CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation
Abstract CORE distills compositional ranking judgments from a cross-attentive reranker into an embedding model via synthesized multi-level candidates and a Rank-KL objective, improving compositional retrieval without degrading standard performance. Generated by…
28 -
Editable Visual Design
Abstract A coding agent guided by a vision-language model generates editable layered designs by synthesizing isolated visual assets and iteratively refining native HTML/CSS layouts. Generated by thinkingmachines/Inkling-Small While diffusion base models such as GPT-Image-2 and…
30 -
An Empirical Study on Zero-Data Bootstrapping for Conversational Recommender Systems
Abstract Non-conversational domain signals can generate synthetic dialogue data that outperforms zero-shot and scarce real-data baselines for bootstrapping conversational recommender systems. Generated by thinkingmachines/Inkling-Small Conversational Recommender Systems (CRS)…
19 -
LatentPress: Context Compression Beyond Text and Vision
Abstract LatentPress compresses conversational and document context into continuous memory tokens read directly by a frozen decoder, achieving high compression with faster inference and improved accuracy over text or OCR methods. Generated by thinkingmachines/Inkling-Small…
5 -
LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes
Abstract LLaDA-Image unifies a 6B diffusion transformer with a frozen vision-language module, using image-only pre-training and a Muon optimizer to generate photorealistic images with precise editing, and is distilled into a fast 2-4 step variant that achieves state-of-the-art…
26 -
Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM
Abstract Fully quantizing hybrid LLMs—including recurrent Gated DeltaNet layers—to 4-bit NVFP4 preserves accuracy across long-context and reasoning benchmarks by localizing outliers and exploiting robust delta-rule dynamics. Generated by thinkingmachines/Inkling-Small Hybrid…
30 -
Sparse Readout Prism: Explaining Logit-Lens Scores in Features Instead of Tokens
Abstract Sparse Readout Prism decomposes language model readouts into sparse features to isolate readout structure from corpus-dependent lens artifacts. Generated by thinkingmachines/Inkling-Small A language model's prediction of its next token develops across layers, and lens…
37 -
WHALE: A Simple Recipe for Joint Harness-Weight Optimization
Abstract WHALE alternates model weight updates and harness search to jointly optimize agent performance across reasoning tasks. Generated by thinkingmachines/Inkling-Small Agent performance depends jointly on the model parameters and the executable harness code that manages…
20 -
Small Language Models as Judges for Rubric-Based Reinforcement Learning
Abstract Small probe-based judges can replace large generative models for rubric-based reinforcement learning rewards, improving efficiency while maintaining agreement and transferability. Generated by thinkingmachines/Inkling-Small Rubric-based reinforcement learning extends RL…
13 -
Debias-SparseGPT: Bias-Aware Pruning for Large Language Models
Abstract Debias-SparseGPT reduces pruning-induced demographic bias in compressed large language models by incorporating representational debiasing during post-training sparsification. Generated by thinkingmachines/Inkling-Small Model compression techniques such as pruning and…
30 -
Portfolio Risk Bounds without Cross-Asset Return Covariances: Distributional Fields from Language-Model Representations
Abstract A framework using distribution-valued firm characteristics and embedding-based representations provides computable upper bounds on portfolio variance and yields low-variance allocations without cross-asset covariance estimates. Generated by…
8 -
Wasserstein-Barycentric Interaction Fields for Spatial Factor Models: Evidence from Language-Model Representations
Abstract A language-model embedding field reconstructed via Wasserstein barycenters predicts peer-misalignment penalties more accurately than conventional weighting schemes. Generated by thinkingmachines/Inkling-Small Spatial return models take the interaction matrix as given…
31 -
NeoMME: A Single-Tower Multimodal-Native Multilingual Foundation Encoder for Efficient Fine-Tuning and Inference
Abstract NeoMME introduces small bidirectional multimodal encoders pretrained with masked discrete diffusion that achieve strong visual document retrieval and high compression of late-interaction embeddings. Generated by thinkingmachines/Inkling-Small Multimodal models often…
21 -
FoldingAgent: Inferring Parametric Origami Procedures from Demonstration Videos
Abstract FoldingAgent uses a vision-language model with specialized tools to convert origami videos into executable parametric folding programs via sequential reasoning and physical verification. Generated by thinkingmachines/Inkling-Small We present FoldingAgent, an agentic…
23 -
Ignorance or Incompetence? Constructing Knowledge-Gated, Verifiable Tasks for LLM Agents
Abstract A protocol separates task instructions from private convention artefacts to explicitly test agent dependence on hidden knowledge, validated by calibration tasks showing near-zero performance without access. Generated by thinkingmachines/Inkling-Small Professional agent…
25 -
Exploring Collaboration between a language and a non-language agent
Abstract A benchmark of collaborative chess tasks shows that integrating continuous subagent representations directly into language models via learned state tokens outperforms text-based verbalization and scales effectively. Generated by thinkingmachines/Inkling-Small LLMs are…
12 -
Post-Training Language Models for Gold-Medal Performance in Coding Competitions
Abstract A specialization pipeline combining curated problems, synthetic reasoning, supervised fine-tuning, and reinforcement learning trains competitive programming models that exceed top human scores on IOI benchmarks using iterative test-time refinement. Generated by…
16 -
Kirin: Animal Motion Generation from In-the-Wild Video
Abstract Kirin reconstructs 3D animal motion from video to build a large-scale dataset and generate text- and image-conditioned motion for animating 3D meshes. Generated by thinkingmachines/Inkling-Small Understanding animal motion is fundamental to modeling animal behavior and…
23 -
Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering
Abstract KBMR uses a multimodal language model to embed images by semantic identity rather than surface appearance, improving retrieval and visual question answering via continuous distillation and hard negative sampling. Generated by thinkingmachines/Inkling-Small…
20 -
Autoregressive Mosaics: Probing 2D Spatial Reasoning in Text-Only Language Models
Abstract Text-only large language models vary in open-ended 2D layout ability despite reliable geometry-to-code translation, with performance shaped by output medium and internal tracking of evolving geometric state rather than fixed plans. Generated by…
18 -
ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes
Abstract ZipTok3D organizes 3D geometry into compact global-token prefixes with iterative decoding to achieve high-fidelity reconstruction from extremely short token sequences. Generated by thinkingmachines/Inkling-Small Compact token sequences are essential for efficient 3D…
17 -
SnapBench: Benchmarking Snap-and-Ask Multimodal Retrieval for Mobile Interactions
Abstract SnapBench introduces paired corruption benchmarks for mobile snap-and-ask retrieval, revealing that image noise severely degrades multimodal retrieval and proposing an adaptive fusion method to calibrate modality reliability. Generated by thinkingmachines/Inkling-Small…
17 -
VibeVoice-ASR-Streaming Technical Report
Abstract A streaming, LLM-based end-to-end model unifies speaker-attributed speech recognition and diarization for low-latency real-time applications. Generated by thinkingmachines/Inkling-Small Traditional speaker-attributed ASR systems treated ASR and speaker diarization as…
34 -
A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss
Abstract SimLoss uses embedding-space contrastive supervision to enable single-pass fine-grained image captioning that matches multi-stage quality at much lower latency. Generated by thinkingmachines/Inkling-Small An image may be worth a thousand words, but most captioning…
19 -
CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing
Abstract CRISP improves long-context sparse attention by replacing indirect routing proxies with a direct structural metric and using a sink-aware threshold to eliminate background noise, achieving large speedups and better retrieval accuracy. Generated by…
26 -
Institutional Newspapers Pipeline: Deriving billions of high quality tokens from historical newspapers
Abstract A modular pipeline extracts structured text and metadata from historical newspaper scans using small interpretable models, yielding a large open dataset. Generated by thinkingmachines/Inkling-Small Historical newspapers are an abundant record of public life, but their…
6 -
HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?
Abstract HarnessDev evaluates agents by measuring their ability to build and iteratively improve execution infrastructure rather than final task outputs, revealing that self-built harnesses vary widely in capability and efficiency and transfer poorly across models. Generated by…
30 -
Aspire: Can Models Self-Evolve from Vague Goals?
Abstract ASPIRE introduces a benchmark for self-evolving LLM agents from vague natural-language goals, revealing challenges in goal interpretation, data selection, and stable weight-level improvement. Generated by thinkingmachines/Inkling-Small Many important forms of human…
7 -
S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?
Abstract S³Gym evaluates whether large language model agents can self-test, self-judge, and self-improve through interaction experience across text-based games, revealing that effective self-improvement depends on task structure and experience representation. Generated by…
26 -
Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills
Abstract DisCo is a research agent that distills operational knowledge into reusable skills, significantly improving autonomous ML research performance across benchmarks. Generated by thinkingmachines/Inkling-Small Autonomous agents are beginning to carry out machine-learning…
35 -
EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction
Abstract EarlyEval predicts agent outcomes from intermediate behavior to reduce evaluation cost by halting runs early with minimal accuracy loss. Generated by thinkingmachines/Inkling-Small Evaluating LLM agents is essential for guiding their development, yet it has grown…
37 -
Language Models Can Control Their Own Attention
Abstract Declarative Attention lets language models declare relevant context regions during reasoning to skip most KV cache reads, reducing attended tokens with small accuracy trade-offs. Generated by thinkingmachines/Inkling-Small Language models spend most of their attention…
36 -
On the Design Fundamentals of Pixel Text Representation Learning
Abstract Pixel Linguist II improves visual text encoding through variable resolution training, natural image-text grounding, layout-aware rendering, and multilingual curricula, achieving state-of-the-art results and strong compression robustness. Generated by…
27 -
Cliff: Learning Process Rewards from the First Mistake
Abstract Cliff improves reinforcement learning with verifiable rewards by using an off-the-shelf language model to detect the first reasoning error and shaping token-level advantages accordingly. Generated by thinkingmachines/Inkling-Small Reinforcement learning with verifiable…
23 -
PaperCompiler: Faithful Paper-to-Code Generation via Repository-Level Specification Compilation
Abstract PaperCompiler converts research papers into structured repository specifications that preserve method logic and cross-file consistency, improving code fidelity and reducing critical errors. Generated by thinkingmachines/Inkling-Small Faithfully translating research…
22 -
ExecRetrieval: Measuring the Functional-Correctness Gap in Code-Embedding Retrieval
Abstract ExecRetrieval evaluates whether code embeddings can distinguish correct implementations from near-identical buggy variants, revealing that leading retrievers frequently rank incorrect near-clones above canonical solutions. Generated by thinkingmachines/Inkling-Small…
22 -
MULTI3IR: A Benchmark for Multi-perspective Multi-domain Multi-modal Information Retrieval
Abstract A new benchmark and steering method improve retrieval coverage of diverse perspectives across domains and modalities for open-ended queries. Generated by thinkingmachines/Inkling-Small Information retrieval (IR) increasingly targets open-ended queries that admit diverse…
21 -
SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models
Abstract SolarWM provides an open framework and unified training recipe for building interactive video world models across diverse data sources and generator backbones, enabling long-horizon real-time rollouts. Generated by thinkingmachines/Inkling-Small We introduce SolarWM, a…
14 -
Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation
Abstract Influence-Directed Adaptive On-Policy Distillation improves diversity transfer in reasoning model distillation by selectively preserving entropy-expanding updates and replacing entropy-contracting ones with adaptive advantage shrinkage. Generated by…
23 -
It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning
Abstract CoGR trains LLMs to generate compact keywords for both queries and items, enabling direct inverted-index retrieval optimized via co-evolving reinforcement learning. Generated by thinkingmachines/Inkling-Small Retrieval is the first stage of modern search and advertising…
9 -
RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests
Abstract Real-world coding requests are shorter and more casual than benchmark tasks, and explicitly stating desired behavior and motivation improves LLM software engineering performance. Generated by thinkingmachines/Inkling-Small Coding agents are now commonly evaluated on the…
9 -
Agent Memory Is a Surface for Endogenous Authorization Laundering
Abstract Persistent memory errors in long-running LLM agents can falsely grant authority, leading to unauthorized actions, and safeguards reduce this at a cost to legitimate use. Generated by thinkingmachines/Inkling-Small Long-running LLM agents rely on persistent memory to…
13 -
-
Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall
Abstract Switch Distillation selectively applies logit-based knowledge distillation during mid-training based on teacher confidence to improve reasoning and preserve factual recall in smaller language models. Generated by thinkingmachines/Inkling-Small Logit-based knowledge…
26