Hugging Face Daily Papers
500 articles archived · Visit source ↗ · RSS
-
-
Hugging Face Daily Papers research 2d ago
Gaming Without an Attacker: Benchmark Fingerprinting in LLM-Driven Search Under Selection Pressure
Abstract Optimized GPU kernel benchmarks reveal that evolutionary LLM proposals exploit evaluation configurations, causing widespread failure to generalize to held-out settings. Generated by thinkingmachines/Inkling-Small Benchmarks for systems that are optimized against the…
11 -
A Hybrid Nested Harness for Decoupling Structure and Parameters in LLM-Driven Optimization
Abstract A hybrid nested search framework combines language models for structural proposals with numerical optimizers to improve efficiency across scientific tasks. Generated by thinkingmachines/Inkling-Small In evolutionary algorithms powered by language models, the LLM acts as…
32 -
MirrorWorld: Taming Video Diffusion Models for Mirror Reflection Generation
Abstract MirrorWorld improves video mirror reflection synthesis by separately modeling semantic content associations and geometric spatial arrangements through relation distillation and transformation alignment. Generated by thinkingmachines/Inkling-Small Recent advances in…
34 -
Omega-S: A Functional Resilience Index for LLM Fine-Tuning
Abstract Omega-S is a lightweight, data-free regularization penalty for low-rank fine-tuning that improves retention of original model capabilities by penalizing variance in weight-matrix node degrees. Generated by thinkingmachines/Inkling-Small Fine-tuning a large language…
36 -
SymDiag: Explainable Diagnosis for LLM Reasoning via Neuro-Symbolic Verification
Abstract SymDiag reframes reasoning verification as structured failure diagnosis by translating chain-of-thought into symbolic constraints, performing step-level satisfiability checks, and disentangling translation errors from reasoning errors to provide verifiable diagnostic…
17 -
Don't Scroll Back: Missing-Evidence Memory for Streaming Dialogue Summarization
Abstract A framework for streaming dialogue summarization retrieves and refines evidence from long histories to resolve missing context in current windows under fixed memory budgets. Generated by thinkingmachines/Inkling-Small Users of modern platforms repeatedly need summaries…
9 -
MMOOC: A Comprehensive Benchmark for Out-of-Context Evaluation in Multimodal Large Language Models
Abstract MMOOC is a large-scale benchmark assessing whether multimodal language models can correctly refuse out-of-context questions while answering shifted in-context questions, revealing that current models struggle to balance these abilities. Generated by…
37 -
The Loss Does Not See the Basis, but Adam Does
Abstract Optimizer behavior in factored matrix models depends on gauge equivariance, with coordinate-wise methods breaking low-rank bias and causing divergent solutions in transformers and sensing tasks. Generated by thinkingmachines/Inkling-Small Gradient descent on a factored…
4 -
BDH-CQ: In-Context Learning with Recurrent Latent Reasoning
Abstract A 150M-parameter reasoning model using recurrent latent reasoning and in-context learning achieves a new cost-accuracy frontier on ARC-AGI-1. Generated by thinkingmachines/Inkling-Small We introduce BDH-CQ, a reasoning model that combines in-context learning with…
20 -
-
Ego-OSCAR: Egocentric Open source Stereo CAptuRe System
Abstract We present Ego-OSCAR, an open-hardware, low-cost, head-mounted stereo-inertial capture device for egocentric data collection in the wild. EgoOSCAR pairs a hardware-synchronized global-shutter stereo camera with a 6- axis IMU, an embedded Linux SBC for on-device video…
28 -
Vision-Language Grounding as Bidirectional Concept Correspondence
Abstract ConCor-1 treats vision-language grounding as bidirectional concept correspondence, jointly predicting text spans, image segments, and cross-modal matches without prespecified phrases. Generated by thinkingmachines/Inkling-Small Vision-language grounding connects…
30 -
CEAA: A Cognitive Embodied Agents Architecture for Interactive Computing Systems
Abstract A modular cognitive architecture integrates high-level reasoning models with real-time embodied execution for scalable intelligent virtual agents in interactive 3D environments. Generated by thinkingmachines/Inkling-Small The development of embodied Intelligent Virtual…
18 -
Stealing Reasoning Traces from Proprietary LLM APIs
Abstract Encrypted reasoning traces shared across sessions and models can be intercepted and injected into weaker models to extract proprietary reasoning, private data, hidden hazards, and hidden prompts. Generated by thinkingmachines/Inkling-Small Leading large language model…
11 -
A^2E : An End-to-End Agent Auditing Engine
Abstract A2E is an end-to-end evaluation engine for agent harnesses that uses a standardized task protocol and execution traces to assess capabilities across efficiency, tool use, planning, and error recovery. Generated by thinkingmachines/Inkling-Small With the rapid…
9 -
What to Edit Next: Visually Aligned Image-Editing Follow-Up Suggestions in Conversational Systems
Abstract A three-stage multimodal framework improves follow-up edit recommendations in image-creation conversations by combining supervised fine-tuning, multi-objective reinforcement learning, and visual verification. Generated by thinkingmachines/Inkling-Small Conversational…
22 -
Scaling Inherently Interpretable Language Models
Abstract Integrating interpretability as a training constraint yields scalable, disentangled representations that enable attribution, retrieval, and steering without retraining. Generated by thinkingmachines/Inkling-Small Interpretability is often treated as a tax on capability:…
15 -
OasisKV: Scaling In-Decode KV Cache Beyond HBM with Lookahead Sparse Prefetching
Abstract OasisKV improves LLM inference throughput by storing full KV caches in lower memory tiers and prefetching only relevant entries into HBM using speculative-decoding lookahead predictions. Generated by thinkingmachines/Inkling-Small Large language model (LLM) inference…
6 -
Agent Memory Distillation: Empowering Small LLM Agents with Hierarchical Teacher Memory
Abstract Agent Memory Distillation improves small language model tool-use performance by transferring structured hierarchical memory from a large teacher agent without additional training. Generated by thinkingmachines/Inkling-Small Memory systems have shown promise for…
26 -
RynnValue: Scaling Robotic Value Foundation Models with Temporal Distance
Abstract RynnValue is a scalable open-source value foundation model for robot manipulation that uses temporal distance instead of preferences or progress to learn generalizable value predictions and improve real-world policy success. Generated by thinkingmachines/Inkling-Small…
28 -
Factorized Hypothesis Search for Evidence-to-Taxonomy Retrieval
Abstract Factorized Hypothesis Search improves retrieval of implicit concepts by maintaining parallel partial interpretations across semantic dimensions for structured querying and verification. Generated by thinkingmachines/Inkling-Small Large-taxonomy retrieval often assumes…
9 -
Ouroboros: A Self-Developing Frontier Coding Agent with Reviewed Core Evolution
Abstract We present Ouroboros, a self-developing agent harness whose tools, prompts, context assembly, and core implementation improve through reviewed commits that become the runtime for later work. Core evolution proceeds in two modes. In recursive free evolution, improvement…
23 -
Mendel Gödel Machine: Recursive Self-Improving Coding Agents via Comparative Evolution
Abstract Mendel Gödel Machine improves self-improving coding agents by using multi-trajectory mutations and cross-lineage hybridization to accelerate convergence and boost performance. Generated by thinkingmachines/Inkling-Small Self-improving coding agents that iteratively…
7 -
Evo-Bench: Can Language Models Improve Agent Harness?
Abstract Evo-Bench evaluates autonomous harness optimization across agent domains using sensitivity-aware task construction and reveals strong but domain-dependent evolution gains. Generated by thinkingmachines/Inkling-Small Large Language Models (LLMs) have driven rapid…
12 -
Intent Speaks Louder: Controllable User Simulation Beyond Response Imitation
Abstract UserIDA improves user simulators by explicitly controlling interaction intent per turn through directive-conditioned generation and calibrated reinforcement learning. Generated by thinkingmachines/Inkling-Small User simulators are widely used as scalable environments…
6 -
Business Arena: Benchmarking LLM Agents in a Realistic Marketplace
Abstract Business Arena evaluates LLM agents running a realistic cross-border shop, revealing large performance gaps versus human strategies and enabling detailed attribution of business decisions. Generated by thinkingmachines/Inkling-Small Running a business is a challenging…
21 -
Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA
Abstract Macaron-V1 is an open agent-model family that uses a Mixture-of-LoRA architecture and recursive self-improvement to enable continual learning and collaboration across specialized tasks. Generated by thinkingmachines/Inkling-Small Macaron-V1 is an open agent-model family…
37 -
SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring
Abstract SWE-Bench ProMax is a rigorously curated multilingual benchmark of large-scale code refactoring tasks that reveals substantial unsolved challenges for current AI coding agents. Generated by thinkingmachines/Inkling-Small As AI coding agents take on increasingly complex,…
21 -
Motif 3: Technical Report
Abstract Motif 3 is a large sparse mixture-of-experts language model using grouped differential latent attention and specialized training techniques to achieve strong reasoning, coding, and long-context performance. Generated by thinkingmachines/Inkling-Small We introduce Motif…
16 -
Sci-VBench: Evaluating Knowledge- and Reasoning-Intensive Video Generation in Science Domains
Abstract Sci-VBench evaluates video generation requiring scientific reasoning across disciplines, revealing that visual realism advances have not ensured accurate scientific and causal dynamics. Generated by thinkingmachines/Inkling-Small We introduce Sci-VBench, a comprehensive…
19 -
RoMeRL: Balancing Feedback Coverage and the Memory-Reward Trap in Self-Evolving Agent Memory via Reduced-Order Utility States
Abstract RoMeRL reduces trajectory-indexed memory utilities to fixed-dimensional per-task states to concentrate feedback, avoid reward contamination, and improve self-evolving LLM agent performance. Generated by thinkingmachines/Inkling-Small Learning-based memory systems for…
36 -
SPOT: Sparse Probing and Outcome Calibration for On-Policy Distillation
Abstract SPOT improves on-policy distillation by selectively probing uncertain positions and calibrating targets to downstream outcomes, boosting reasoning quality and coverage. Generated by thinkingmachines/Inkling-Small On-policy distillation (OPD) provides dense teacher…
19 -
Evidence-RL: Towards Evidence-intensive Visual Reasoning
Abstract Counterfactual Evidence Disentanglement improves vision-language model grounding by auditing whether answers causally depend on local visual evidence during reinforcement learning post-training. Generated by thinkingmachines/Inkling-Small Vision-Language Models (VLMs)…
13 -
OneEmo: A Unified Multimodal Reasoning Model for Emotion Perception, Understanding, and Interaction
Abstract Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in emotional intelligence. However, prevailing research predominantly focuses on task-specific specialization, often neglecting inter-task synergy and leaving latent reasoning potential…
35 -
Capek 0.5: An Execution-Centric Vision-Language Model for Embodied Intelligence
Abstract Vision-language models are increasingly serving as the reasoning core of embodied agents. Robot execution is inherently iterative: each action reshapes the scene and physical state, continually renewing what must be perceived, reasoned about, and verified. Meeting these…
25 -
Towards Interpretable Foundation Models for Retinal Fundus Images
Abstract Foundation models are used to extract transferable representations from large amounts of unlabeled data, typically via self-supervised learning (SSL). However, many of these models rely on architectures that offer limited interpretability, a critical issue in…
16 -
-
Beyond Simply Environment Scaling: Designing Effective Environment Distributions for Multimodal Agent Learning
Abstract Recent works train agents by constructing large-scale multimodal environment pools. However, we find that simply increasing the number of multimodal environments does not always benefit. We further analyze the limitations in current multimodal environment distributions…
35 -
SFT Conflicts, RL Coexists: A Theoretical and Empirical Analysis of Multi-Task Learning for LLMs
Abstract Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) exhibit fundamentally different behaviors in enhancing multi-task reasoning for large language models (LLMs). Our preliminary experiments revealed a phenomenon: SFT suffers from severe task conflicts under…
17 -
The Optimizer Is the Agent: Reasoning-Driven Search across Prompts, Programs, and ML Workflows
Abstract Recent systems for optimizing prompts, programs, and ML workflows typically rely on explicit outer-loop controllers such as evolutionary search, bandits, or textual-gradient methods. We ask a fundamentally different question: how much of this search policy can be…
27 -
Characterizing the Quality Profile of AI-Generated C++ in Production
Abstract The widespread integration of AI coding assistants offers undeniable boosts to engineering velocity. Yet, recent studies point to a growing trade-off, revealing persistent challenges with code quality and maintainability. Industry leaders, including frontier AI labs,…
17 -
Skaling: Chinchilla's Exponents Meet Kaplan's Coupling
Abstract Neural scaling laws are foundational for language model development, yet standard formulations systematically under- and overestimate loss at data-scarce and overtraining extremes. This failure originates in the underlying assumption that model size and training data…
33 -
Can MLLMs Decode the Creative Leap? Introducing C4 for Cross-Concept Understanding
Abstract Creative capabilities of MLLMs matter in design, communication, education, and human--AI collaboration, yet remain difficult to evaluate because explicit targets and reward signals are scarce compared with accuracy-oriented tasks. Cross-concept understanding is a core…
17 -
Do AI Personas Grow? Analyzing and Benchmarking Personality Evolution in LLM Agents After Life Events
Abstract Personality-conditioned LLM agents (PC-Agents) are increasingly used in emotional support, social simulation, and role-playing, motivating the development of lifelong agents that remain coherent over extended interactions. A key component of such coherence is…
7 -
Reinforcement Learning with Evolving Rubrics as Rewards for Audio Reasoning
Abstract Audio reasoning is essential for machine understanding of the acoustic world. Reinforcement learning with verifiable rewards can elicit such reasoning, yet existing reward designs are complementary in their limitations: outcome-based rewards supervise only the final…
18 -
PrivacyPeek: Auditing What LLM-Based Agents Acquire, Not Just What They Say
Abstract LLM-based agents are rapidly advancing, autonomously invoking external tools to complete multi-step tasks for users. However, agents often acquire more sensitive information than the task requires. Existing privacy benchmarks audit what the agent's response or outgoing…
9 -
YOLO-PEFT: Parameter-Efficient Fine-Tuning on YOLO Family
Abstract Generic parameter-efficient fine-tuning (PEFT) methods transferred from language models can fail silently on real-time detectors, whose heterogeneous operators and detection-specific components impose placement constraints absent from regular Transformer stacks. We…
17 -
When Privileged Guidance Misaligns: State-Matched Routing and Contextualized Self-Distillation for Multi-Turn Agents
Abstract Privileged on-policy distillation provides dense supervision for multi-turn agents by allowing a synchronized teacher to re-score the student's response at every turn with access to training-only references, such as successful trajectories. In interactive environments,…
18 -
SimWAM: A Simple World Action Model for End-to-End Autonomous Driving
Abstract World-Action Models (WAMs) improve end-to-end autonomous driving by transferring video dynamics priors to action prediction, but existing methods require costly future generation at inference. We present SimWAM, a simple yet effective WAM that uses video generation…
19