Search: “llm”

When Do Large Language Models Exhibit Unsolicited Deception?

arXiv:2504.00285v2 Announce Type: replace Abstract: Large Language Models (LLMs) are effective at deceiving when prompted to do so. Models that demonstrate better performance on reasoning tasks are al

agentsarxiv-cs-cl

10 Sept 2026

Who Are They to Each Other? Multi-Agent Reasoning for Speaker Relationship Inference

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

DGX agent

arXiv:2609.09628v1 Announce Type: cross Abstract: Inferring speaker relationships from spoken conversations is an important step towards socially aware speech understanding. However, this task remains

agentsarxiv-cs-cl

10 Sept 2026

Who Maintains Agent Skills? A Longitudinal Study of Human-Governed, AI-Assisted Skill Maintenance

arXiv:2609.05677v1 Announce Type: cross Abstract: Lifelong LLM agents increasingly rely on external skill artifacts as one element for preserving and reusing capabilities over time. These skills (usua

agentsarxiv-cs-ai

10 Sept 2026

A Closed-Form Estimator and Diagnostic Battery for Anchor-Judge Error Correlation, Under a Single-Common-Factor Model

arXiv:2609.08826v1 Announce Type: cross Abstract: When an external reference set (an anchor) is used to decompose an LLM-judge panel's error into a quality signal and a shared common-mode error, stand

safetyarxiv-cs-lg

CrossModalQA: A Cross-modal and Multi-hop Benchmark for Multimodal Retrieval-augmented Generation

model-releasesarxiv-cs-cv

arXiv:2609.05518v1 Announce Type: new Abstract: Despite the strong capabilities of multimodal large language models (MLLMs), their parametric knowledge remains incomplete and difficult to update, moti

From Narrative to Auditable Forecasts: A Structured Scaffold for Agentic Forecasting

arXiv:2609.05905v1 Announce Type: cross Abstract: LLM agents are increasingly used for live forecasting, where they retrieve up-to-date information and produce estimates for unresolved future events.

agentsarxiv-cs-lg

I like this framing a lot: agent improvement is harness improvement, not model improvement The interesting interventions are often at the to…

I like this framing a lot: agent improvement is harness improvement, not model improvement The interesting interventions are often at the tool boundary: - what context you pass - when tools are provid

agentsharrison-chase--x

MOAE: Multi-Objective Agent Evolution with Pareto-Preserving Search

model-releasesarxiv-cs-lg

arXiv:2609.05992v1 Announce Type: cross Abstract: As LLM-based agents continue to advance, their evaluation has become increasingly multifaceted: a capable agent must not only achieve high task comple

RePro: Training Language Models to Faithfully Recycle the Web for Pretraining

arXiv:2510.10681v2 Announce Type: replace-cross Abstract: High-quality data is a cornerstone of large language model (LLM) pretraining, yet its growth has not kept pace with the needs of frontier mode

researcharxiv-cs-lg

Rethinking the Evaluation of Efficiency Methods for Multi-Agent Systems

model-releasesarxiv-cs-lg

arXiv:2609.05933v1 Announce Type: new Abstract: Efficiency is increasingly important for Large Language Model (LLM)-based multi-agent systems (MAS), as larger models and more agents introduce substant

Why CPUs Are the Workhorse of Agentic AI Infrastructure

While GPUs remain essential for LLM training and inference, enterprise‑scale agentic AI workloads shift the focus to orchestration rather than raw compute. Agents must break down requests into subtask

agentsvultr

Local Ai

An open-source context layer for building AI on top of company data

We’ve been building PipesHub for a while now, and I’d love to get more teams to try it and tell us where it breaks. The problem we kept running into was pretty simple: Building an AI app over company

local-air-localllama

8 Sept 2026

I reduced image-processing token usage by ~95% compared with GPT-4o direct vision, while maintaining roughly the same accuracy.How significant is that?[P]

model-releasesr-machinelearning

I'm testing a new approach for reducing the cost of image-based LLM inference. I evaluated it on the MOMA Graph benchmark, using 1,315 questions. Compared with using GPT-4o to process the original ima

8 Sept 2026

au^au-Bench: An Environment for End-To-End, Realistic Agent Construction

arXiv:2609.04611v1 Announce Type: new Abstract: LLM agents are rapidly becoming production software, deployed to handle customer service, adjudicate disputes, and operate internal systems. Notably, th

Can Activation Steering Capture Multidimensional Authorship Style?

arXiv:2609.04792v1 Announce Type: cross Abstract: Activation steering has shown promise for controlling LLM generation along well-defined attributes, but it remains unclear whether it can handle the m

CoSkill: Joint Reinforcement Learning of Reasoning and Meta-Skill Agents for Hierarchical Skill Evolution

arXiv:2609.04865v1 Announce Type: new Abstract: Skill libraries improve the sample efficiency of agentic reinforcement learning (RL) by enabling large language model (LLM) agents to reuse procedural k

safetyarxiv-cs-ai

EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?

arXiv:2609.04280v1 Announce Type: cross Abstract: Modern LLM-based agents operate through a harness of tools, reusable skills, and specialist agents that shapes what they observe and what they can do.

How to Speculate about Uncertainty in Agentic Coding? A Draft-Model Gate Method

model-releasesarxiv-cs-lg

arXiv:2609.05274v1 Announce Type: new Abstract: LLM agents deployed for software engineering fail expensively: they act confidently wrong, and bad actions are recognized only after costly execution an

KV cache as an agent runtime [R]

Our research team has been exploring an alternative approach to achieving interactivity and better responsiveness with LLM systems. One of the team members wrote up a post about it: https://research.y

agentsr-machinelearning

Large Language Models for HVAC Operations in Building Energy Systems: A Critical Review of Methods, Applications, and Deployment Readiness

arXiv:2609.05314v1 Announce Type: new Abstract: Building automation systems generate rich sensor data yet remain insight-poor because heterogeneous point naming, missing metadata, and fragmented docum

safetyarxiv-cs-ai

Large Language Models with At Most One Spike per Neuron

arXiv:2609.05151v1 Announce Type: cross Abstract: Leveraging their inherent sparse event-driven computation, spiking neural networks (SNNs) offer a promising path toward energy-efficient large languag

researcharxiv-cs-cl

Leveraging Low-Level Symbolic Competences for Unsupervised Grounding in Hallucination Detection

arXiv:2609.05025v1 Announce Type: cross Abstract: Hallucination-where a language model generates outputs that are factually incorrect or unsupported by the source-is a major challenge for both prompte

Multi-Step Tool-Calling over Korean Open Public APIs: A Benchmark and a Data-Synthesis Recipe

arXiv:2609.05395v1 Announce Type: new Abstract: Data-sovereignty regulations increasingly require public institutions to deploy open-source, on-premise LLM agents that chain multiple tool-calls across

Persistent Teacher Anchoring for Tool-Using Agents

arXiv:2609.04773v1 Announce Type: cross Abstract: Distillation is common in LLM post-training, where on-policy knowledge distillation (OPKD) uses student-generated trajectories to prepare the student

safetyarxiv-cs-ai

Qwen3.8-27B on M1 Max 32GB: MLX 15.8 tok/s vs llama.cpp 9.7 tok/s - but llama.cpp prefill is faster

I’ve been setting up an M1 Max Mac Studio (24-core GPU, 32GB unified memory) as a local LLM server and wanted to compare MLX vs llama.cpp on Qwen3.8-27B. I tried to keep the model footprint and benchm

model-releasesr-ollama

Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective

arXiv:2609.04489v1 Announce Type: cross Abstract: Pause-token methods improve LLM reasoning by inserting special tokens into sequences. Prior work explains these gains through computational expressivi

Uncertainty Signals for Network Intent Translation: Risk Ranking and Ambiguity Localization

arXiv:2609.04486v1 Announce Type: cross Abstract: Intent-based networking realization starts by translating high-level intents into low-level network configurations. Recent approaches have shifted tow

Planning to get a cheap-ish GPU. Would appreciate some advice.

model-releasesr-localllama

Hi, I've been wanting to run my own local LLM for some time and I finally saved enough to get a budget GPU. I can spend 700 at most, and I'm looking for a GPU that can run quantized 30B~ parameter mod

6 Sept 2026

Adapting to Evolving Requirements: Agentic AI for Retail Supply Chain Operations

arXiv:2609.03860v1 Announce Type: new Abstract: Retail supply chain operations rely on coupled decision modules that must adapt as requirements evolve. LLMs offer a natural-language interface for this

Beyond BLEU: A Case for Redefining Sign Language Translation Benchmarks

arXiv:2609.03734v1 Announce Type: cross Abstract: BLEU-4 is the standard metric for evaluating sign language translation (SLT), but spoken-language metrics may not adequately reflect sign language pro

IndicSafeEval: Safety Robustness of Large Language Models under Multilingual Persuasive Jailbreak Attacks

arXiv:2609.03781v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used in multilingual settings, yet their safety is still evaluated primarily in English. This limits our

Learning What Not to Forget: Long-Horizon Agent Memory from a Few Kilobytes of Learning

arXiv:2606.20954v2 Announce Type: replace-cross Abstract: Long-running language-model systems accumulate interaction history that outgrows the context window, so they must continually evict. When an e

Margins, Not Windows: Training-Free Per-Step Lossy Speculative Decoding

arXiv:2609.02897v1 Announce Type: new Abstract: Speculative decoding accelerates LLM inference by drafting candidate tokens and verifying them in parallel. Tree-attention drafters such as EAGLE-3 are

SWIM: Student Writing Simulation via Proficiency-Conditioned Generation

arXiv:2609.03215v1 Announce Type: new Abstract: Writing proficiency manifests in how students develop content, organize ideas, choose words, and use language. Despite growing interest in LLM-based stu

safetyarxiv-cs-cl

The Civilization Framework: Sovereign-Anchored Communication Between Personal Multi-Agent Systems

arXiv:2609.03425v1 Announce Type: cross Abstract: Humans are the transport layer between AI systems, losing context at every hop. We present the Civilization Framework, whose addressable party is the

agentsarxiv-cs-ai

What Do CAE Simulation Agents Really Need Beyond a Generic Harness?

arXiv:2609.03718v1 Announce Type: cross Abstract: Computer-aided engineering (CAE) simulation is among the largest and most demanding areas of engineering, where setting up a solver such as OpenFOAM,

agentsarxiv-cs-cl

What Else Needs Fixing? Exploring Cost-Effective Test-Time Compute for Revision Propagation in Artifacts Generated Through Conversation

arXiv:2609.03254v1 Announce Type: new Abstract: Large Language Models (LLMs) often help users generate artifacts through iterative cycles of generation and revision in conversation. A challenge here i

You Can't Escape Your Own Activations : Evaluation Awareness and Multi-Agent Monitoring

arXiv:2609.03035v1 Announce Type: cross Abstract: LLM agents are increasingly deployed in multi-agent systems, where they can collude while keeping their actions benign. Output monitors designed to de

agentsarxiv-cs-lg

Are Non-English Papers Reviewed Fairly? Language-of-Study Bias in NLP Peer Reviews

arXiv:2604.07119v2 Announce Type: replace Abstract: Peer review plays a central role in the NLP publication process, but is susceptible to various biases. Here, we study language-of-study (LoS) bias:

safetyarxiv-cs-cl

CHIME: Credit-Aware Hierarchical Memory Evolution for Long-Horizon Agentic Planning

arXiv:2609.02074v1 Announce Type: new Abstract: Planning is a central capability that enables agents to decompose complex long-horizon tasks into manageable steps. Test-time search and training-based

CHisAgent: A Multi-Agent Framework for Event Taxonomy Construction in Ancient Chinese Cultural Systems

arXiv:2601.05520v2 Announce Type: replace Abstract: Despite strong performance on many tasks, large language models (LLMs) show limited ability in historical and cultural reasoning, particularly in no

safetyarxiv-cs-cl

Local Ai

Coverage, Not Targeting: A Structural Regime in Multi-Turn Agent Credit Assignment

arXiv:2609.02417v1 Announce Type: cross Abstract: Multi-turn agentic RL increasingly treats credit assignment as a targeting problem: given a terminal verifiable reward, per-turn methods localize cred

local-aiarxiv-cs-ai

CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing

arXiv:2609.01925v1 Announce Type: cross Abstract: The attention prefilling phase of long-context LLM inference scales quadratically, making self-attention a severe computational bottleneck. Traditiona

researcharxiv-cs-cl

Do Large Language Models Capture the Diversity in their Training Data?

arXiv:2609.02275v1 Announce Type: cross Abstract: Large language models are trained to model conditional distributions over text, yet it remains inadequately understood whether they capture the full d

EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction

arXiv:2609.02783v1 Announce Type: new Abstract: Evaluating LLM agents is essential for guiding their development, yet it has grown prohibitively expensive: a single pass of a frontier model over an ag

Tutorials

EmoStance: Response-Side Affective-Orientation Control for Empathetic Response Generation via Emoji Weak Supervision

arXiv:2609.02133v1 Announce Type: new Abstract: Empathetic response generation requires models to decide not only what to say, but also how to respond to the previous speaker's affective situation. We

tutorialsarxiv-cs-ai

Gauge dependence and structured-output corruption in sign-branched repetition penalties: measurements across models, inference stacks, and alternative repetition controls

arXiv:2607.09791v2 Announce Type: replace-cross Abstract: The multiplicative repetition penalty shipped across the LLM inference ecosystem (HuggingFace, vLLM, llama.cpp, and a dozen further engines) b

HyperStyler: Low-resource Authorship Style Transfer via Context-aware Style Navigation and Hypernetworks

arXiv:2609.02772v1 Announce Type: new Abstract: Low-resource authorship style transfer (LAST) aims to rewrite text into the style of an arbitrary target author using only a few reference examples whil