Knowledge catalogue
arXiv:2605.14163v1 Announce Type: new Abstract: Can a committee of weak reasoning-model calls reach the performance of much stronger models? We study verifier-backed committee search as inference-time
arXiv:2509.26100v2 Announce Type: replace Abstract: The rapid integration of Large Language Models (LLMs) into high-stakes domains necessitates reliable safety and compliance evaluation. However, exis
arXiv:2506.08584v4 Announce Type: replace Abstract: Medical question answering (QA) benchmarks often focus on multiple-choice or fact-based tasks, leaving open-ended answers to real patient questions
arXiv:2605.14068v1 Announce Type: new Abstract: We introduce CurveBench, a benchmark for hierarchical topological reasoning from visual input. CurveBench consists of extbf{756 images} of pairwise non-
arXiv:2509.23023v3 Announce Type: replace Abstract: Large language models are increasingly deployed in multi-agent settings whose outcomes hinge on social intelligence, motivating evaluations of their
arXiv:2508.01916v3 Announce Type: replace-cross Abstract: Understanding internal representations of neural models is a core interest of mechanistic interpretability. Due to its large dimensionality, t
arXiv:2605.14403v1 Announce Type: new Abstract: Dermatological diagnosis requires integrating fine-grained visual perception with expert clinical knowledge. Although Multimodal Large Language Models (
arXiv:2605.14912v1 Announce Type: new Abstract: Pluralistic alignment is typically operationalised as preference aggregation: producing responses that span (Overton), steer toward (Steerable), or prop
arXiv:2605.14790v1 Announce Type: cross Abstract: Research idea generation is the innovation-driving step of automated scientific research. Recently, large language models (LLMs) have shown potential
i appreciate how seriously the team always takes these reports (even when the answer turns out to be 'i got used to the current level of magic and now i'd like more please') Codex team is aware of rep
arXiv:2511.15407v3 Announce Type: replace Abstract: Humans learn by observing, interacting with environments, and internalizing physics and causality. Here, we aim to ask whether an agent can similarl
arXiv:2605.14087v1 Announce Type: new Abstract: Large Language Models (LLMs), when trained on web-scale corpora, inherently absorb toxic patterns from their training data. This leads to ``toxic degene
arXiv:2605.06132v2 Announce Type: replace Abstract: In agent memory systems, the reranking model serves as the critical bridge connecting user queries with long-term memory. Most systems adopt the 're
arXiv:2605.14635v1 Announce Type: cross Abstract: This paper introduces a multi-label visual emotion analysis benchmark dataset for comprehensively evaluating the ability of multimodal large language
arXiv:2602.21545v3 Announce Type: replace Abstract: Muon has recently emerged as a strong optimizer for large language model pre-training, orthogonalizing the momentum matrix via Newton--Schulz polar
arXiv:2605.15071v1 Announce Type: cross Abstract: Vision-Language Models (VLMs) are increasingly applied to cultural heritage materials, from digital archives to educational platforms. This work ident
arXiv:2507.07776v3 Announce Type: replace Abstract: Unrestricted adversarial attacks aim to fool computer vision models without being constrained by ell_p-norm bounds to remain imperceptible to humans
arXiv:2605.14051v1 Announce Type: new Abstract: Industrial LLM agent systems often separate planning from execution, yet LLM planners frequently produce structurally invalid or unnecessarily long work
arXiv:2510.13016v3 Announce Type: replace Abstract: A truly capable AI system must do more than detect objects or recognize activities in isolation. It must form unified, grounded representations of w
arXiv:2605.14604v1 Announce Type: new Abstract: This position paper argues that effective tutoring requires corrective friction: surfacing misconceptions and challenging them supportively to drive con
arXiv:2605.15053v1 Announce Type: cross Abstract: Continually pre-training a large language model on heterogeneous text domains, without replay or task labels, has remained an unsolved architectural p
arXiv:2605.14091v1 Announce Type: new Abstract: In recent years, the rapid evolution of generative AI has fundamentally reshaped the paradigm of image forgery, breaking the traditional boundaries betw
arXiv:2605.14542v1 Announce Type: new Abstract: A skilled live-commerce host is not merely a narrator, but a sales agent who converts viewer curiosity into purchase intent through expert product knowl
arXiv:2605.14478v1 Announce Type: cross Abstract: Context: Retrieval-augmented code generation relies on cross-file repository context, but retrieved snippets may come from obsolete project states. Ob
arXiv:2605.14504v1 Announce Type: new Abstract: Long-horizon household tasks demand robust high-level planning and sustained reasoning capabilities, which are largely overlooked by existing embodied A
arXiv:2409.02038v3 Announce Type: replace-cross Abstract: Existing text-to-SQL benchmarks have largely been constructed from public databases with well-structured schemas and simplistic question-SQL p
arXiv:2605.13415v1 Announce Type: cross Abstract: This paper presents a multi-stage framework for detecting reclaimed slurs in multilingual social media discourse. It addresses the challenge of identi
arXiv:2603.26839v2 Announce Type: replace-cross Abstract: How do multimodal models solve visual spatial tasks -- through genuine planning, or through brute-force search in token space? We introduce ex
arXiv:2605.13536v1 Announce Type: cross Abstract: High-Level Synthesis (HLS) compiles algorithmic C/C++ descriptions into hardware, with Quality of Results (QoR) -- latency and resource utilization --
arXiv:2605.13773v1 Announce Type: cross Abstract: Large Language Models (LLMs) are being employed widely to automate tasks across the software development life-cycle. It is, however, unclear whether t
arXiv:2605.12724v1 Announce Type: cross Abstract: Instruction-based image editing exhibits heterogeneous difficulty not only across cases but also across regions of an image, motivating refinement app
arXiv:2603.03295v2 Announce Type: replace-cross Abstract: Whether in agentic workflows, social studies, or chat settings, large language models (LLMs) are increasingly being asked to replace humans in
arXiv:2505.15616v2 Announce Type: replace Abstract: Multimodal Large Language Models (MLLMs) have achieved significant advances in integrating visual and linguistic information, yet their ability to r
arXiv:2605.13829v1 Announce Type: cross Abstract: We introduce Negation Neglect, where finetuning LLMs on documents that flag a claim as false makes them believe the claim is true. For example, models
arXiv:2601.09636v2 Announce Type: replace Abstract: While GUI agents have shown strong performance under explicit and completion instructions, real-world deployment requires aligning with users' more
arXiv:2603.20334v2 Announce Type: replace-cross Abstract: In high-complexity abstract reasoning, a system must infer a latent rule from a few examples or structured observations and apply it to unseen
arXiv:2605.13695v1 Announce Type: cross Abstract: LLM-as-a-judge is now the default measurement instrument for open-ended generation, but on the public JudgeBench benchmark even strong instruction-tun
arXiv:2602.07458v4 Announce Type: replace Abstract: Online Reinforcement Learning (RL) offers a promising avenue for complex image editing but is currently constrained by the scarcity of reliable and
arXiv:2605.12817v1 Announce Type: cross Abstract: Longitudinal clinical notes contain rich evidence of how patients evolve over time, but converting this signal into training supervision for clinical
arXiv:2605.13835v1 Announce Type: new Abstract: Class-Incremental Learning (CIL) enables models to continuously integrate new knowledge while mitigating catastrophic forgetting. Driven by the remarkab
arXiv:2605.12978v1 Announce Type: new Abstract: Learning from past experience benefits from two complementary forms of memory: episodic traces -- raw trajectories of what happened -- and consolidated
arXiv:2510.15297v4 Announce Type: replace-cross Abstract: We introduce VERA-MH (Validation of Ethical and Responsible AI in Mental Health), an automated evaluation of the safety of AI chatbots used in
Asymmetry of capabilities is when there is great risk. Diffusion of capabilities maintains an adversarial equilibrium and stability. Don't let the fear-mongerers win. We asked the CEO of HuggingFace @
arXiv:2605.12271v1 Announce Type: new Abstract: Humans often specify and create through visual artifacts: typography sheets, sketches, reference images, and annotated scenes. Yet modern visual generat
arXiv:2605.11143v1 Announce Type: new Abstract: Reasoning benchmarks measure clinical performance on clean inputs. We evaluate the step before reasoning: retrieval over real EHR notes, where negation,
Co-sign. We asked the CEO of HuggingFace @ClementDelangue what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. 'Six, seven years ago, at
Tool: CSP Allow-list Experiment An experiment that shows that you can load an app in a CSP-protected sandboxed iframe (see previous note) and have a custom fetch() that intercepts CSP errors and passe
arXiv:2605.11086v1 Announce Type: cross Abstract: AI agents are rapidly gaining capabilities that could significantly reshape cybersecurity, making rigorous evaluation urgent. A critical capability is
Good catch @theojaffee 😂 Media We asked the CEO of HuggingFace @ClementDelangue what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. 'Si
arXiv:2605.11209v1 Announce Type: new Abstract: While existing benchmarks demonstrate the near-perfect performance of large language models (LLMs) on various tasks, this apparent saturation often obsc
arXiv:2605.11181v1 Announce Type: new Abstract: The recent empirical success of the Muon optimizer has renewed interest in non-Euclidean optimization, typically justified by similarities with second-o
arXiv:2605.11396v1 Announce Type: new Abstract: The Muon optimizer has emerged as a compelling alternative to Adam for training large language models, achieving remarkable computational savings throug
arXiv:2605.11730v1 Announce Type: new Abstract: Automated red-teaming for LLMs often discovers narrow attack slices, missing diverse real-world threats, and yielding insufficient data for safety fine-
arXiv:2605.12422v1 Announce Type: new Abstract: Automatic generation of educational materials using large language models (LLMs) is becoming increasingly common, but assigning difficulty levels to suc
arXiv:2605.11907v1 Announce Type: new Abstract: We measure procedural-skill SFT contribution across three Qwen3.5 dense scales (0.8B, 2B, 4B) on a 200-task / 40-skill holdout, with Claude Haiku 4.5 as
arXiv:2605.11680v1 Announce Type: new Abstract: We introduce ShapeCodeBench, a synthetic benchmark for perception-to-program reconstruction: given a rendered raster image, a model must emit an executa
arXiv:2605.12466v1 Announce Type: cross Abstract: Looped Transformers offer a promising alternative to purely feed-forward computation by iteratively refining latent representations, improving languag
arXiv:2605.11574v1 Announce Type: new Abstract: The literature on how large language models handle conflict between their training knowledge and a contradicting document presents a persistent empirica
We asked the CEO of HuggingFace @ClementDelangue what the risks of releasing powerful open source models are. He says restricting AI creates more risk than openness. 'Six, seven years ago, at the time
arXiv:2605.08734v1 Announce Type: cross Abstract: Low-Rank Adaptation (LoRA) reparameterizes a weight update as a product of two low-rank factors, but the Jacobian J_{G} of the generator mapping the f