Search: “llm”

Meta-Aligner: Bidirectional Preference-Policy Optimization for Multi-Objective LLMs Alignment

arXiv:2604.24178v1 Announce Type: cross Abstract: Multi-Objective Alignment aims to align Large Language Models (LLMs) with diverse and often conflicting human values by optimizing multiple objectives

28 Apr 2026

Agents

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

PARASITE: Conditional System Prompt Poisoning to Hijack LLMs

DGX agent

arXiv:2505.16888v4 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are increasingly deployed via third-party system prompts downloaded from public marketplaces. We identify a criti

agentsarxiv-cs-ai

28 Apr 2026

BLAST: Benchmarking LLMs with ASP-based Structured Testing

arXiv:2604.22306v1 Announce Type: cross Abstract: Large Language Models (LLMs) have demonstrated remarkable performance across a broad spectrum of tasks, including natural language understanding, dial

Agents

For the past few years, humans have been doing “prompt engineering” to coax the best performance out of different LLMs. In this work, we exp…

For the past few years, humans have been doing “prompt engineering” to coax the best performance out of different LLMs. In this work, we explored what happens if we train an AI to do that job instead.

agentsdavid-ha--x

LLMPhy: Parameter-Identifiable Physical Reasoning Combining Large Language Models and Physics Engines

arXiv:2411.08027v3 Announce Type: replace-cross Abstract: Most learning-based approaches to complex physical reasoning sidestep the crucial problem of parameter identification (e.g., mass, friction) t

NiuTrans.LMT: Toward Inclusive and Scalable Multilingual Machine Translation with LLMs

arXiv:2511.07003v2 Announce Type: replace Abstract: Large language models have significantly advanced Multilingual Machine Translation (MMT), yet scaling to many languages while keeping quality robust

Removing Sandbagging in LLMs by Training with Weak Supervision

arXiv:2604.22082v1 Announce Type: cross Abstract: As AI systems begin to automate complex tasks, supervision increasingly relies on weaker models or limited human oversight that cannot fully verify ou

When Cow Urine Cures Constipation on YouTube: Limits of LLMs in Detecting Culture-specific Health Misinformation

arXiv:2604.22002v1 Announce Type: new Abstract: Social media platforms have become primary channels for health information in the Global South. Using gomutra (cow urine) discourse on YouTube in India

1. I never said LLMs were not useful. They are, particularly with all the bells and whistles that are being added to them. I use them. 2. A …

1. I never said LLMs were not useful. They are, particularly with all the bells and whistles that are being added to them. I use them. 2. A robot-rich future can't be built with AIs that don't underst

researchyann-lecun--x

And eventually schools will (if they have any sense) probably withdraw or limit LLMs.

And eventually schools will (if they have any sense) probably withdraw or limit LLMs. 🦔Schools across the US are reversing years of technology-first classroom policies after studies show laptop and sc

safetygary-marcus--x

Are LLMs really more important than fire or electricity? “Honestly, a ton of what we’ve developed in my lifetime amounts to scaling up the d…

Are LLMs really more important than fire or electricity? “Honestly, a ton of what we’ve developed in my lifetime amounts to scaling up the delivery of information and entertainment and the frictionles

safetygary-marcus--x

Breaking Bad: Interpretability-Based Safety Audits of State-of-the-Art LLMs

model-releasesarxiv-cs-lg

arXiv:2604.20945v1 Announce Type: cross Abstract: Effective safety auditing of large language models (LLMs) demands tools that go beyond black-box probing and systematically uncover vulnerabilities ro

Local Ai

DeepAgents Sandbox — a lightweight native Linux sandbox for deepagents It gives full isolation for safely running LLM/agent-generated code —…

local-aiharrison-chase--x

DeepAgents Sandbox — a lightweight native Linux sandbox for deepagents It gives full isolation for safely running LLM/agent-generated code — no Docker, no VMs, no cloud dependency. Super fast and secu

Applications

Reasoning Primitives in Hybrid and Non-Hybrid LLMs

arXiv:2604.21454v1 Announce Type: cross Abstract: Reasoning in large language models is often treated as a monolithic capability, but its observed gains may arise from more basic operations. We study

applicationsarxiv-cs-ai

SafeRedirect: Defeating Internal Safety Collapse via Task-Completion Redirection in Frontier LLMs

arXiv:2604.20930v1 Announce Type: cross Abstract: Internal Safety Collapse (ISC) is a failure mode in which frontier LLMs, when executing legitimate professional tasks whose correct completion structu

StructMem: Structured Memory for Long-Horizon Behavior in LLMs

arXiv:2604.21748v1 Announce Type: cross Abstract: Long-term conversational agents need memory systems that capture relationships between events, not merely isolated facts, to support temporal reasonin

Automatic Ontology Construction Using LLMs as an External Layer of Memory, Verification, and Planning for Hybrid Intelligent Systems

arXiv:2604.20795v1 Announce Type: new Abstract: This paper presents a hybrid architecture for intelligent systems in which large language models (LLMs) are extended with an external ontological memory

Avoiding Overthinking and Underthinking: Curriculum-Aware Budget Scheduling for LLMs

arXiv:2604.19780v1 Announce Type: new Abstract: Scaling test-time compute via extended reasoning has become a key paradigm for improving the capabilities of large language models (LLMs). However, exis

Can 'AI' Be a Doctor? A Study of Empathy, Readability, and Alignment in Clinical LLMs

arXiv:2604.20791v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly deployed in healthcare, yet their communicative alignment with clinical standards remains insufficiently

CCTVBench: Contrastive Consistency Traffic VideoQA Benchmark for Multimodal LLMs

model-releasesarxiv-cs-cv

arXiv:2604.20460v1 Announce Type: new Abstract: Safety-critical traffic reasoning requires contrastive consistency: models must detect true hazards when an accident occurs, and reliably reject plausib

Diagnosing CFG Interpretation in LLMs

arXiv:2604.20811v1 Announce Type: new Abstract: As LLMs are increasingly integrated into agentic systems, they must adhere to dynamically defined, machine-interpretable interfaces. We evaluate LLMs as

Do you want to know why LLMs feel sharp on surface semantics but hollow on the fine-grained stuff? “From Tokens to Thoughts: How LLMs and Hu…

Do you want to know why LLMs feel sharp on surface semantics but hollow on the fine-grained stuff? “From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning” Come by the poster tomor

researchyann-lecun--x

DRIV-EX: Counterfactual Explanations for Driving LLMs

arXiv:2603.00696v2 Announce Type: replace Abstract: Large language models (LLMs) are increasingly used as reasoning engines in autonomous driving, yet their decision-making remains opaque. We propose

HiPO: Hierarchical Preference Optimization for Adaptive Reasoning in LLMs

arXiv:2604.20140v1 Announce Type: new Abstract: Direct Preference Optimization (DPO) is an effective framework for aligning large language models with human preferences, but it struggles with complex

Hybrid Policy Distillation for LLMs

arXiv:2604.20244v1 Announce Type: cross Abstract: Knowledge distillation (KD) is a powerful paradigm for compressing large language models (LLMs), whose effectiveness depends on intertwined choices of

Memorization, Emergence, and Explaining Reversal Failures: A Controlled Study of Relational Semantics in LLMs

arXiv:2601.02931v2 Announce Type: replace Abstract: Autoregressive LLMs perform well on relational tasks that require linking entities via relational words (e.g., father/son, friend), but it is unclea

DP-FlogTinyLLM: Differentially private federated log anomaly detection using Tiny LLMs

arXiv:2604.19118v1 Announce Type: cross Abstract: Modern distributed systems generate massive volumes of log data that are critical for detecting anomalies and cyber threats. However, in real world se

Easy Samples Are All You Need: Self-Evolving LLMs via Data-Efficient Reinforcement Learning

arXiv:2604.18639v1 Announce Type: cross Abstract: Previous LLMs-based RL studies typically follow either supervised learning with high annotation costs, or unsupervised paradigms using voting or entro

Improving the Distributional Alignment of LLMs using Supervision

arXiv:2507.00439v4 Announce Type: replace Abstract: The ability to accurately align LLMs with diverse population groups on subjective questions would have great value. In this work, we show that addin

Let me say this clearly: LLMs cannot feel emotions. Emotions are evolutionary mechanisms. They push us to avoid danger or approach what is b…

Let me say this clearly: LLMs cannot feel emotions. Emotions are evolutionary mechanisms. They push us to avoid danger or approach what is beneficial. We experience emotions because we are alive, and

safetygary-marcus--x

LLMs Know They're Wrong and Agree Anyway: The Shared Sycophancy-Lying Circuit

arXiv:2604.19117v1 Announce Type: new Abstract: When a language model agrees with a user's false belief, is it failing to detect the error, or noticing and agreeing anyway? We show the latter. Across

safetyarxiv-cs-lg

Agents

Mind the (DH) Gap! A Contrast in Risky Choices Between Reasoning and Conversational LLMs

arXiv:2602.15173v2 Announce Type: replace Abstract: The use of large language models either as decision support systems, or in agentic workflows, is rapidly transforming the digital ecosystem. However

agentsarxiv-cs-ai

Multi-modal Reasoning with LLMs for Visual Semantic Arithmetic

arXiv:2604.19567v1 Announce Type: new Abstract: Reinforcement learning (RL) as post-training is crucial for enhancing the reasoning ability of large language models (LLMs) in coding and math. However,

STReasoner: Empowering LLMs for Spatio-Temporal Reasoning in Time Series via Spatial-Aware Reinforcement Learning

arXiv:2601.03248v2 Announce Type: replace Abstract: Spatio-temporal reasoning in time series involves the explicit synthesis of temporal dynamics, spatial dependencies, and textual context. This capab

ArgBench: Benchmarking LLMs on Computational Argumentation Tasks

arXiv:2604.17366v1 Announce Type: new Abstract: Argumentation skills are an essential toolkit for large language models (LLMs). These skills are crucial in various use cases, including self-reflection

DART: Mitigating Harm Drift in Difference-Aware LLMs via Distill-Audit-Repair Training

arXiv:2604.16845v1 Announce Type: new Abstract: Large language models (LLMs) tuned for safety often avoid acknowledging demographic differences, even when such acknowledgment is factually correct (e.g

Do LLMs Use Cultural Knowledge Without Being Told? A Multilingual Evaluation of Implicit Pragmatic Adaptation

arXiv:2604.17718v1 Announce Type: new Abstract: Many benchmarks show that large language models can answer direct questions about culture. We study a different question: do they also change how they s

Evaluating Multimodal LLMs for Inpatient Diagnosis: Real-World Performance, Safety, and Cost Across Ten Frontier Models

model-releasesarxiv-cs-lg

arXiv:2604.16980v1 Announce Type: new Abstract: Background: Large language models (LLMs) are increasingly proposed for diagnostic support, but few evaluations use real-world multimodal inpatient data,

Expressing Social Emotions: Misalignment Between LLMs and Human Cultural Emotion Norms

arXiv:2604.16757v1 Announce Type: new Abstract: The expression of emotions that serve social purposes, such as asserting independence or fostering interdependence, is central to human interactions and

FreezeEmpath: Efficient Training for Empathetic Spoken Chatbots with Frozen LLMs

arXiv:2604.18159v1 Announce Type: new Abstract: Empathy is essential for fostering natural interactions in spoken dialogue systems, as it enables machines to recognize the emotional tone of human spee

From Fallback to Frontline: When Can LLMs be Superior Annotators of Human Perspectives?

arXiv:2604.17968v1 Announce Type: cross Abstract: Although large language models (LLMs) are increasingly used as annotators at scale, they are typically treated as a pragmatic fallback rather than a f

HPLT 3.0: Very Large-Scale Multilingual Resources for LLMs and MT. Mono- and Bi-lingual Data, Multilingual Evaluation, and Pre-Trained Models

arXiv:2511.01066v3 Announce Type: replace Abstract: We present an ongoing initiative to provide open, very large, high-quality, and richly annotated textual datasets for almost 200 languages. At 30 tr

Applications

Large Language Models Are Bad Dice Players: LLMs Struggle to Generate Random Numbers from Statistical Distributions

arXiv:2601.05414v2 Announce Type: replace Abstract: As large language models (LLMs) transition from chat interfaces to integral components of stochastic pipelines and systems approaching general intel

applicationsarxiv-cs-cl

LLMs are still not consistent judges of qualitative work, and small changes to how that work is presented affect outcomes. Better harnessing…

model-releasesethan-mollick--x

LLMs are still not consistent judges of qualitative work, and small changes to how that work is presented affect outcomes. Better harnessing and methods (multiple judging runs with randomized orders,

MedRedFlag: Investigating how LLMs Redirect Misconceptions in Real-World Health Communication

arXiv:2601.09853v2 Announce Type: replace Abstract: Real-world health questions from patients often unintentionally embed false assumptions or premises. In such cases, safe medical communication typic

No One Fits All: From Fixed Prompting to Learned Routing in Multilingual LLMs

arXiv:2604.16937v1 Announce Type: new Abstract: Translation-based prompting is widely used in multilingual LLMs, yet its effectiveness varies across languages and tasks. We evaluate prompting strategi

Applications

Rethinking Uncertainty Estimation in LLMs: A Principled Single-Sequence Measure

arXiv:2412.15176v3 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly employed in real-world applications, driving the need to evaluate the trustworthiness of their generat

applicationsarxiv-cs-lg

StealthGraph: Exposing Domain-Specific Risks in LLMs through Knowledge-Graph-Guided Harmful Prompt Generation

arXiv:2601.04740v3 Announce Type: replace Abstract: Large language models (LLMs) are increasingly applied in specialized domains such as finance and healthcare, where they introduce unique safety risk