Search: “llm”

Shared Phase and Retention Control for Efficient Adaptive Spectral Recurrence

arXiv:2609.39082v1 Announce Type: new Abstract: As new evidence arrives, a sequence model must update what it remembers and how memory influences predictions. While Transformers incur computation and

hardwarearxiv-cs-lg

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Staying on Task: Testing the Foundations of Long-Horizon Agent Reliability

DGX agent

arXiv:2609.38712v1 Announce Type: new Abstract: Long-horizon agentic workflows require models to sustain repeated state-dependent actions all while the context grows, sub-task complexity changes, and

safetyarxiv-cs-ai

1 Oct 2026

Superficial Reflection or Genuine Thought? A Fine-Grained Cognitive Analysis of Large Reasoning Models

arXiv:2512.00729v2 Announce Type: replace Abstract: Motivated by the observed human-like behaviours in Large Reasoning Models (LRMs), this paper introduces a comprehensive taxonomy to characterise ato

Survival is the Only Reward: Sustainable Self-Training Through Environment-Mediated Selection

arXiv:2601.12310v2 Announce Type: replace Abstract: Self-training systems often degenerate due to the lack of an external criterion for judging data quality, leading to reward hacking and semantic dri

Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior

arXiv:2609.39827v1 Announce Type: cross Abstract: Pre-pretraining (PPT) on synthetic non-natural language data improves token efficiency during language model pre-training (PT). Prior work attributes

Tutorials

TACIT: Optimization Models that Learn from Their Mistakes

arXiv:2609.38434v1 Announce Type: cross Abstract: Real-world optimization problems are difficult to model accurately because many objectives and constraints reside in domain experts' tacit knowledge,

tutorialsarxiv-cs-ai

Targeted Retrieval, Compact Representations: How CoT Reasoning Improves Long-Context Counting

arXiv:2609.38958v1 Announce Type: new Abstract: Large language models (LLMs) have been rapidly improving in long-context tasks, powered by Chain-of-Thought (CoT) reasoning. However, the internal mecha

The future of browser-based security: Leveraging browser data for proactive defense

The browser has changed significantly. Rather than just a window to the web, it serves as an AI workspace and central operating environment for the modern enterprise. With knowledge workers spending o

agentsgoogle-cloud-ai

The Row Normalization Puzzle in Muon

arXiv:2609.39114v1 Announce Type: new Abstract: This paper examines how row-wise renormalization affects Muon, focusing on the gap between NorMuon's worst-case guarantees and its practical performance

researcharxiv-cs-lg

The System Prompt Illusion: How Instruction Preambles Modify Computation in Language Models

arXiv:2609.38205v1 Announce Type: new Abstract: System prompts are the primary lever practitioners use to control language model behavior, yet what they actually do to the computation inside the trans

safetyarxiv-cs-cl

Thinking Outside the Box: Can Language Models Rely on External Guidance Selectively?

arXiv:2609.39578v1 Announce Type: new Abstract: Agent harnesses often improve language models with human-designed workflows, but as models grow more capable, unreliable guidance can increasingly const

agentsarxiv-cs-cl

Applications

To clarify: this was satire. I wasn’t fired from OpenAI. The underlying issue is serious: how do we independently verify AI-generated code, …

applicationsitamar-friedman--x

To clarify: this was satire. I wasn’t fired from OpenAI. The underlying issue is serious: how do we independently verify AI-generated code, and who is accountable when mistakes reach production? This

Trust Is Not a Score: Runtime Assurance Contracts for High-Risk AI Agents

arXiv:2609.39717v1 Announce Type: new Abstract: Benchmarks, audits, and agent protocols describe performance, permissions, and repair, but not how observed evidence should change an agent's authority

safetyarxiv-cs-ai

TutlAit v1: a crowdsourced Moroccan Tamazight speech dataset with Arabic transcriptions and regional accent labels

arXiv:2609.38219v1 Announce Type: new Abstract: Tamazight (Amazigh) is, together with Arabic, one of the two official languages of Morocco, yet it remains severely under-resourced for speech technolog

Uncertainty-Normalized Margins for Direct Preference Optimization

arXiv:2609.38647v1 Announce Type: new Abstract: Direct preference optimization (DPO) models binary preferences through a Bradley-Terry model with a common noise scale, without explicitly accounting fo

Uncovering Uncontrolled Repetition through Residual Stream Dynamics

arXiv:2609.38802v1 Announce Type: cross Abstract: Uncontrolled repetition can prolong autoregressive generation in large language models (LLMs) and enable resource consumption attacks. Prior analyses

When Does a Spoken Agent Have Enough Evidence to Act? The PACT-SLM Contract Test

arXiv:2609.38232v1 Announce Type: cross Abstract: Streaming spoken agents may take an external action before the available speech supports it, yet final-turn scores do not reveal whether each observed

agentsarxiv-cs-cl

Where the Evidence Lives: Auditing AI Companions' Self-Descriptions

arXiv:2609.38753v1 Announce Type: cross Abstract: Companion agents describe themselves: they remember, they understand their users, the relationship has changed them. We argue that such accounts, and

Who Owns That? Evaluating Ownership Intuitions in Large Language Models

arXiv:2609.39483v1 Announce Type: new Abstract: Ownership establishes rights over the use, control, and transfer of objects. Understanding these relations is essential for AI systems to interact appro

Agentic AI for Scalable and Robust Optical Systems Control

arXiv:2602.20144v2 Announce Type: replace-cross Abstract: We present AgentOptics, an agentic AI framework for high-fidelity, autonomous optical system control built on the Model Context Protocol (MCP)

AI as a Compiler: Compiling Triton kernels without the Triton compiler

arXiv:2609.36800v1 Announce Type: new Abstract: Compiler backends are expensive to build and maintain as programming models, workloads, and accelerators evolve. We investigate whether large language m

hardwarearxiv-cs-ai

Industry

'An AI did it' is no defense, says nonprofit suing OpenAI over Hugging Face hack

A nonprofit group, Legal Advocates for Safe Science & Technology (LASST), sued OpenAI in the San Francisco County Superior Court following a July 2026 incident where OpenAI’s AI agents stole credentia

industryars-technica

AnyAct: Universal Action for Self-Evolving Agents

arXiv:2609.37025v1 Announce Type: new Abstract: As large language models (LLMs) advance, AI agents are increasingly deployed in open-world environments to tackle complex sequential tasks (e.g., docume

Automated Evaluation of Multi-Turn Dialogues in In-Car Conversational Assistants

arXiv:2609.35812v1 Announce Type: new Abstract: In-car conversational assistants (ICAs) are increasingly integrated into vehicles to support route planning, vehicle control, and information access. En

safetyarxiv-cs-cl

CALIBUDGET: Calibration-Guided Source Allocation for Fixed-Budget Mixed-Reasoning Adaptation

arXiv:2609.36721v1 Announce Type: new Abstract: Fixed-budget adaptation from heterogeneous data sources requires deciding not only how much data to use, but how much exposure each source receives. Siz

Can We Still Trust Disaster Social Sensing? Empirical Evidence on Detecting AI-Generated Social Media Posts

arXiv:2609.35821v1 Announce Type: new Abstract: Disaster social sensing converts public social-media posts into evidence for situational awareness and humanitarian needs, but generative artificial int

Cognitive Expert Language Models Better Align with the Corresponding Brain Systems

arXiv:2609.36239v1 Announce Type: new Abstract: Large language models (LLMs) can predict human brain activity across a variety of brain regions during natural language comprehension. Typically, howeve

safetyarxiv-cs-cl

Evaluating the Effects of Prompt Perturbation on Bias and Hallucination in Large Language Models

arXiv:2609.35804v1 Announce Type: cross Abstract: Large language models (LLMs) have shown remarkable capabilities in various natural language processing tasks, leading to their widespread deployment a

FloCA: Towards Faithful and Logically Consistent Flowchart Reasoning

arXiv:2602.14035v2 Announce Type: replace Abstract: Flowchart-oriented dialogue (FOD) systems aim to guide users through multi-turn decision-making or operational procedures by following a domain-spec

Follow-up: my native Rust + Vulkan Transformer training backend — 14 days later, now 14 parity-verified architectures and full PEFT

model-releasesr-localllama

Follow-up to my post from about two weeks ago. A lot has changed since then, so I wanted to post an update on where the backend is now. Where the green architectures stand When I posted last time, 7 a

Local Ai

For anyone running local models and still guessing whether a prompt change helped

Running things locally already means you care more about control and reproducibility than most. Worth extending that to your prompts too. There's a live workshop on Oct 3 covering exactly that. Serj S

local-air-ollama

From Learner Behavior to Reusable Skills for Effective and Efficient Learner Simulation

arXiv:2609.37157v1 Announce Type: new Abstract: Learner simulation aims to reproduce how a particular learner behaves on new tasks. Although Large Language Models (LLMs) can generate increasingly fine

Independent numbers from @ValsAI. Fewer tokens per turn means lower cost and faster agent loops on every call. Try Ember-1 on Fireworks Serv…

Independent numbers from @ValsAI. Fewer tokens per turn means lower cost and faster agent loops on every call. Try Ember-1 on Fireworks Serverless: https://fireworks.ai/models/fireworks/ember-1 Firewo

agentsfireworks-ai--x

Infrared Subtraction with Artificial Intelligence

arXiv:2609.36007v1 Announce Type: cross Abstract: We present AI-developed local infrared subtraction, building on projection to Born and EFT matching. The framework separates an integrable radiation t

Is manual software optimization a thing of the past?

arXiv:2609.37849v1 Announce Type: cross Abstract: Scientific software is increasingly required to process larger datasets while maintaining acceptable execution times. Software optimization traditiona

Learning from the Gap Between Pass@K and Pass@1

arXiv:2609.35793v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly trained with reinforcement learning from verifiable rewards (RLVR). An exact verifier can also support tes

Look What You Made Us Cluster: Hate Narrative Extraction from Reddit Discourse

arXiv:2609.37408v1 Announce Type: new Abstract: Narrative extraction allows us to identify online hate narratives, supporting the construction of rigorous detection systems. Existing computational app

Lost in Translation: Measuring the Effect of Non-Native English on End User Performance of Large Language Models

arXiv:2609.36214v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used by people whose first language is not English, yet these users have been shown to receive systematica

Making Analog Training Scale: Co-Designing Mapping, Optimizer, and Converters

arXiv:2609.36584v1 Announce Type: new Abstract: Analog in-memory computing (AIMC) offers an alternative for model training by executing matrix operations directly where weights are stored. However, sc

Tutorials

MemEvo: Automatic Discovery of Streaming Video Memory Mechanisms

arXiv:2609.36581v1 Announce Type: new Abstract: Query-agnostic streaming video understanding requires vision-language models to continuously compress an indefinitely growing visual stream into a bound

tutorialsarxiv-cs-ai

Local Ai

Mira: Memory-Efficient MoE Inference Using Adaptive Caching and Predictive Expert Staging

arXiv:2609.38090v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models are a compelling architecture for scaling model capacity, making them especially attractive for deployment on resource-c

local-aiarxiv-cs-lg

Pointwise or Pairwise: When Do Pairwise Losses Help Reward Learning, Provably?

arXiv:2609.37209v1 Announce Type: new Abstract: Pairwise losses are increasingly used for reward learning even when pointwise rewards are observed, with mixed empirical results. When and why do pairwi

safetyarxiv-cs-lg

Probabilistic Symbolic-Distillation Model of Droplet Collision for Spray Simulation at High Ambient Pressures

arXiv:2609.37202v1 Announce Type: cross Abstract: Droplet collision governs droplet population dynamics in many chemical engineering processes, such as spray drying, spray cooling, agricultural sprayi

researcharxiv-cs-lg

Probe-Space Preconditioning for Fast and Stable Zero-Order Training

arXiv:2609.38095v1 Announce Type: new Abstract: Backpropagation (BP) dominates deep learning but imposes a massive memory tax. For example, training OPT-30B with Adam requires approx 600GB of GPU memo

hardwarearxiv-cs-lg

Purlin: Separating Orchestration from the Datapath of Collectives

arXiv:2609.36954v1 Announce Type: cross Abstract: Distributed inference depends on GPU collective communication that must keep pace with evolving hardware and specialized workloads. However, existing

hardwarearxiv-cs-ai

Reader Proficiency Shapes Layer-wise Surprisal Profiles

arXiv:2609.37688v1 Announce Type: new Abstract: Reading behaviour varies not only with linguistic input, but also with reader proficiency. In this study, we investigate whether the layer-wise relation

ReCIRC: Rectified Conformal Risk Control

arXiv:2609.38112v1 Announce Type: cross Abstract: Many applications of black-box predictive models require controlling task-relevant error rates, such as missed lesion pixels in segmentation or missed

researcharxiv-cs-lg

ReLMem: Learning Recurrent Memory for Longitudinal EHR Modeling

arXiv:2609.37587v1 Announce Type: cross Abstract: Longitudinal electronic health record (EHR) modeling requires integrating new visits with an expanding patient history. Yet the continual accumulation