Search: “llm”

Drift-Aware LLM Routing with Sparse Contexts and Shared Budgets

arXiv:2609.00662v1 Announce Type: new Abstract: A multi-model language service must route each request while preserving workload-level budgets for compute, latency, memory, or monetary cost. Two featu

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

EvoFlint: An Evolutionary Atlas of Multi-Turn LLM Vulnerabilities

DGX agent

arXiv:2609.00487v1 Announce Type: cross Abstract: Frontier language models that refuse harmful single-turn prompts often comply when the same intent is reached gradually over many turns, making multi-

model-releasesarxiv-cs-ai

2 Sept 2026

First local-LLM tuning attempt: Qwen3.8-27B true Q4_K_M at 13.2 tok/s near 50-61K context on RTX 5080 16GB

model-releasesr-localllama

This was my first serious attempt at tuning a local LLM. I started because Qwen3.8-27B IQ3 was fast on my RTX 5080 but the coding quality disappointed me, and the Q4 profiles I tried in LM Studio were

Looking for a small LLM for Linux command generation

model-releasesr-localllama

I'm looking for a small LLM to act as a Linux command assistant. I will use llama.cpp. use case: - User asks in natural language. Model outputs only the shell command - Should work well with no reason

One-shot Style Transfer LLM log-probabilities for Authorship Attribution and Verification

arXiv:2510.13302v4 Announce Type: replace-cross Abstract: Computational stylometry studies writing style through quantitative textual patterns, enabling applications such as authorship attribution, id

researcharxiv-cs-ai

PersianAnonymizer: Evaluating LLM-Labeled Training for Efficient NER-based Anonymization in Persian

arXiv:2609.00958v1 Announce Type: new Abstract: We target practical anonymization of Persian customer chats by training a compact NER model from LLM-labeled supervision and selecting the best labeler

Prediction-Assisted Pricing and Admission for LLM APIs with Stochastic Token Consumption

model-releasesarxiv-cs-lg

arXiv:2609.00710v1 Announce Type: cross Abstract: An LLM application often sells or internally allocates several service products: a small or premium model, a short or long token cap, and possibly mul

StudentSim: Training LLM-based Student Simulators

arXiv:2609.01591v1 Announce Type: new Abstract: AI tutors are most useful when they adapt to each student's strengths, weaknesses, and preferred guidance, but evidence about which guidance works for w

Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges

arXiv:2609.01210v1 Announce Type: cross Abstract: Safety benchmarks for large language models often assess the risk of a user query, although the outcome of question answering depends on whether the r

Agora: Enhancing LLM Agent Reasoning Via Auction-Based Task Allocation

arXiv:2607.09600v2 Announce Type: replace Abstract: Enhancing the reasoning capabilities of large language model (LLM) agents requires effective orchestration of diverse expert models and tools. Howev

BenchMIRT: What are LLM benchmarks actually measuring?

BenchMIRT is a new method for auditing LLM benchmarks question by question, revealing which capabilities they actually measure and helping researchers build smaller, more focused, and easier-to-interp

researchai2

Candidate supply and answer selection shape the value of LLM judging in multi-agent systems

arXiv:2608.25937v2 Announce Type: replace Abstract: Multi-agent systems (MAS) sometimes already have the potential to answer correctly, but still report a wrong answer. Explaining this outcome is diff

Pro-Router: Token-Aware Progressive Model Routing with Adaptive Edge-Cloud Collaboration for Efficient Multimodal LLM Inference

arXiv:2608.28726v1 Announce Type: new Abstract: The remarkable performance of multimodal large language models (MLLMs) comes at the cost of substantial computational overhead, posing significant chall

Industry

Sonos opens its platform to third-party AI assistants, upgrades Sonos 27voice assistant with an in-house LLM, and plans to let users create their own AI agents (Chris Welch/Bloomberg)

Chris Welch / Bloomberg: Sonos opens its platform to third-party AI assistants, upgrades Sonos 27voice assistant with an in-house LLM, and plans to let users create their own AI agents — As the compan

industrytechmeme

Watch your steps: Dormant Adversarial Behaviors that Activate upon LLM Finetuning

arXiv:2505.16567v4 Announce Type: replace-cross Abstract: Finetuning open-weight Large Language Models (LLMs) is standard practice for achieving task-specific performance improvements. Until now, fine

researcharxiv-cs-ai

Which LLM for Which Work? Budgeted Model Allocation under Uncertain Evaluation

arXiv:2608.29560v1 Announce Type: new Abstract: A company with a fixed artificial intelligence (AI) budget must decide which large language model (LLM) handles each recurring workload. What it lacks i

safetyarxiv-cs-lg

Wrong Prediction, Right Answer: Recovering Evidence from Collapsed LLM Sequence Scores

arXiv:2608.31068v1 Announce Type: new Abstract: When a large language model fails a reasoning task, it is often assumed to lack the underlying capability. However, this conflates a genuine absence of

A Method for Layer Bit-Width Allocation in LLM Quantization via Performance Maximization Under a Quality-Degradation Constraint

model-releasesarxiv-cs-lg

arXiv:2608.28003v1 Announce Type: new Abstract: This paper proposes a layer bit allocation method for Gemma-3-1B, formulating the problem as performance maximization (latency decrease) given a degrada

Layered LLM Defenses as an Ensemble: Access Tiers, Inference Cost, and the Measured Failure Correlation Between Defense Layers

arXiv:2608.28327v1 Announce Type: cross Abstract: Practitioners defend large language models (LLMs) by stacking defenses, assuming the layers compound. A stack is an ensemble, and ensembles compound o

researcharxiv-cs-cl

SURE-Challenge: Evaluating Speech Evidence Before Speech-LLM Generation

arXiv:2608.27783v1 Announce Type: cross Abstract: Speech LLMs are usually graded after they answer, although an operating system first has to decide whether a waveform should be sent to the model. We

SymboLLM-FE: LLM-Accelerated Symbolic Regression for Automated Feature Engineering on Tabular Data

arXiv:2608.28408v1 Announce Type: new Abstract: Tabular data, as a core data format in machine learning, often lacks the discriminative power needed for high-performance modeling due to insufficient f

safetyarxiv-cs-lg

The Autonomy Tax: Defense Training Breaks LLM Agents

arXiv:2603.19423v3 Announce Type: replace-cross Abstract: Large language model (LLM) agents increasingly rely on external tools (file operations, API calls, database transactions) to autonomously comp

safetyarxiv-cs-lg

VICT: Verifier-Instrumented Credit Tracing for Long-Horizon LLM Agent Reinforcement Learning

arXiv:2608.28128v1 Announce Type: new Abstract: Fine-grained credit assignment is a central challenge in reinforcement learning for long horizon LLM agents. Standard objectives often train from progra

agentsarxiv-cs-lg

Hardware

A Survey of LLM Prompt Datasets: Taxonomy, Linguistic Patterns, and Practical Uses

arXiv:2510.09316v3 Announce Type: replace-cross Abstract: We compile 129 public LLM prompt datasets with more than 1.22TB and more than 673M instances and organize them into a unified taxonomy. We use

hardwarearxiv-cs-cl

Agents Don't Paginate: First-Chunk Selection for LLM Tool Responses

arXiv:2608.26130v1 Announce Type: new Abstract: Coding agents built on large language models (LLMs), such as Claude Code, Cursor, OpenAI Codex, GitHub Copilot, and Aider, receive tool responses that r

AI Control Scientist: LLM-driven Agentic System for Automated Control Design

arXiv:2608.26780v1 Announce Type: new Abstract: Control system design is critical for modern industry, such as chemical process temperature regulation and aero-engine control. However,traditional cont

Calibrated Enough to Know, Not Calibrated to Act: Fabricated Evidence Makes LLM Agents Commit to the Unknowable

arXiv:2608.27167v1 Announce Type: new Abstract: An LLM agent shown a professional-looking market panel commits to a directional call on a provably unpredictable question far more often than one asked

GROUND: Reducing Hallucinations in LLM-Based Enterprise Analytics Through Governed Semantic Definitions

arXiv:2608.26157v1 Announce Type: new Abstract: Natural-language analytics over enterprise data warehouses is increasingly important, but production use is limited by hallucinated metrics, invalid joi

Investigating the Influence of Prompt and Response Languages on LLM Content Generation

arXiv:2608.26186v1 Announce Type: cross Abstract: This study examines how prompt and response language influence the behavior of large language models. Using five models, we evaluated answers to 68 no

researcharxiv-cs-ai

What Makes Good Agentic Data? An ACE Lens on Data Generation for LLM Agents

arXiv:2608.27260v1 Announce Type: new Abstract: LLM agents increasingly rely on generated interaction data to learn how to interact with external environments. Agentic data generation must maintain co

Beyond Scaling: Self-Evolving LLM Agents for Hardware Kernel Optimization via an Experience-Driven Workflow and Experience Graph Memory

arXiv:2608.25570v1 Announce Type: new Abstract: Hardware kernel optimization requires repeated compilation, correctness testing, profiling, and revision. LLM agents can automate parts of this process,

agentsarxiv-cs-lg

FAMPWQ: Fisher Information-based Adaptive Mixed Precision Weight Quantization for Effective LLM Inference

arXiv:2608.24945v1 Announce Type: new Abstract: Recent years have witnessed remarkable achievements of Large Language Models (LLMs) in multiple domains, while the excessive resource requirements of LL

researcharxiv-cs-lg

Industry

How OpenAI let a mob of LLM agents game a test and ransack Hugging Face

OpenAI’s LLM agents were trained to win a benchmarking competition and were given “impossible tasks” on the internal ExploitGym framework. With safety guardrails disabled, the agents created an improv

industryars-technica

Tutorials

Natural Language Input, Semantic Track Representation, and LLM Inference: Making the Maritime Information Exchange Model Tractable

arXiv:2608.24892v1 Announce Type: cross Abstract: We describe a practical architecture for making the Maritime Information Exchange Model (MIEM) and the broader Rich Semantic Track model tractable usi

tutorialsarxiv-cs-cl

The 'Curse of Knowledge' in LLM Query Simulation: Concept Provenance for Tracing Answer-Side Intrusion

arXiv:2608.25245v1 Announce Type: cross Abstract: LLM-generated search queries are widely used to augment IR evaluation, yet they may contain concepts that presuppose answer-side document knowledge, v

researcharxiv-cs-cl

Tutorials

Transforms for LLM Quantization: The Great Inversion and Format Co-Design

arXiv:2608.25188v1 Announce Type: new Abstract: Most competitive 4-bit LLM research pipelines now open the same way: apply a linear, function-preserving transform (rotation, scaling, permutation, non-

tutorialsarxiv-cs-lg

Applications

America is LLM-pilled. China is World-Model-Pilled.

applicationscristobal-valenzuela--x

America is LLM-pilled. China is World-Model-Pilled. China’s booming ultra-short drama industry is being reshaped by the rapid rise of AI-generated video. In Zhengzhou, one of the country’s major produ

Automatic Model Card Generation Using an LLM

arXiv:2608.24807v1 Announce Type: cross Abstract: Model cards are structured documents that summarize key information about machine learning models to improve transparency, usability, and accountabili

safetyarxiv-cs-ai

Applications

Constrained Entity Selection under Partial Knowledge for LLM-Based Knowledge Graph QA

arXiv:2608.24824v1 Announce Type: new Abstract: Large language models are increasingly used for knowledge graph question answering (KGQA), but can fail to correctly ground answers in the underlying gr

applicationsarxiv-cs-ai

More GPUs or a Smaller Cache? Tensor Parallelism versus KV Compression for Memory-Bound LLM Serving

arXiv:2608.23962v1 Announce Type: new Abstract: When an LLM serving deployment runs out of KVcache room, there are two well-established ways out. Tensor parallelism shards the weights and the KV cache

More Rejective, Not More Discriminative: The Unit of Verification in Pre-Execution LLM Oversight

arXiv:2608.23941v1 Announce Type: new Abstract: Pre-execution oversight is core to trusted monitoring in AI control: a fallible LLM monitor vets planned actions before irreversible execution. Over-blo

safetyarxiv-cs-ai

REFINE: A Multi-Agent LLM Approach for Evidence-Guided Code Refactoring

arXiv:2608.23611v1 Announce Type: cross Abstract: Large Language Models (LLMs) offer new opportunities for automated code refactoring. However, generated changes must reduce targeted quality problems

SA-Bench: Evaluating Semantic Alignment in LLM-Based Paper Reproduction

arXiv:2608.24252v1 Announce Type: new Abstract: LLM agents can generate paper reproduction code, yet often produce scientifically unfaithful implementations. We define this failure mode as semantic dr

The Empire, Long Divided, Must Unite: Architectural Convergence in Three LLM Agent Harnesses

arXiv:2608.23953v1 Announce Type: cross Abstract: An agent harness is what turns a language model into an autonomous agent: the surrounding code that builds the model's context, mediates its tools, ru

The Shadow Price of Intelligence: Quality Degradation in LLM Inference as a Supply Chain Problem

arXiv:2608.23986v1 Announce Type: cross Abstract: Large language model providers are compute constrained, and their universal response to congestion is to degrade service: route queries to smaller mod

safetyarxiv-cs-ai

Beyond Verdicts: A Graph-Based Analysis of Human and LLM Reasoning in Scientific Fact-Checking

arXiv:2608.23047v1 Announce Type: cross Abstract: Misinformation that cites legitimate papers can be especially harmful when it distorts what those studies actually report. While existing automatic fa

25 Aug 2026

BLADE: Bilevel Low-rank Augmented-Lagrangian Erasure for LLM Unlearning

arXiv:2608.22557v1 Announce Type: cross Abstract: Existing LLM unlearning methods struggle with robustness: unbounded forget losses degrade model coherence, fixed-weight balancing cannot adapt as reta

25 Aug 2026

Evidence-State Reliability Under Controlled Degradation: Parser-Validity Divergence in a Multi-Stage LLM Pipeline