Search: “gpt”

Interesting safety paper from Microsoft. They find that a weaker, unaligned model can split a harmful task into harmless-looking subquestions, ask an aligned frontier model each one in a separate sess

arXiv:2609.16748v1 Announce Type: new Abstract: Text summarization requires models to condense content while preserving key qualities such as consistency and coherence. Large language models (LLMs) ha

arXiv:2609.14803v1 Announce Type: new Abstract: This paper reports experiments across six frontier model types from OpenAI, Anthropic, xAI, and Google DeepMind. Ten independent sessions per model type

arXiv:2609.14248v1 Announce Type: cross Abstract: Faithful citation attribution begins with identifying the intended source for a scientific claim. We study this source-identification capability throu

arXiv:2609.15849v1 Announce Type: cross Abstract: Can LLMs reason through new information like humans, or do they merely retrieve cached opinions? This is critical for silicon sampling, where LLM pers

arXiv:2609.14352v1 Announce Type: new Abstract: AI-generated image detection has attracted increasing attention, but existing evaluations mainly focus on natural images, leaving AI-generated document

arXiv:2609.14715v1 Announce Type: new Abstract: We scale our conventional sub-150M pretraining recipe from 53.5M to 109.7M parameters, holding the method fixed (Qwen3-style decoder with grouped-query

arXiv:2609.15383v1 Announce Type: cross Abstract: Language model safety is typically evaluated one interaction at a time. We show that a weaker, unaligned model can split a harmful task into benign-lo

arXiv:2609.14839v1 Announce Type: cross Abstract: The integration of Large Language Models (LLMs) into automated code optimization introduces a critical reliability risk we term the Efficiency Halluci

arXiv:2609.13192v1 Announce Type: new Abstract: This study compares traditional machine learning models and Large Language Model (LLM)-generated rule-based systems for heart disease prediction using t

arXiv:2609.13816v1 Announce Type: cross Abstract: JavaScript powers approximately 98.8% of all websites, making vulnerabilities in its code a significant security risk, yet existing detection approach

arXiv:2609.13847v1 Announce Type: new Abstract: In this paper, we extend FLORES+ with Portuguese-source evaluation sets for three Mozambican Bantu varieties: Xichangana, Mozambican Nyanja, and Sena. W

arXiv:2609.13841v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used for emotional support and relationship advice, where a model's tendency to preserve a user's face can

arXiv:2603.18614v2 Announce Type: replace Abstract: Tool-augmented large language models (LLMs) must tightly couple multi-step reasoning with external actions, yet existing benchmarks often confound t

arXiv:2609.12162v1 Announce Type: cross Abstract: Draft-verify-revise is a common LLM orchestration pattern for scaling inference-time compute. One LLM drafts, a second critiques the draft and provide

arXiv:2609.12331v1 Announce Type: new Abstract: Simulated students generated by computational models provide a practical way to evaluate tutoring strategies and pedagogical approaches used by human an

Does anyone have any idea how this could vaguely be possible given the nature of and various routes to RSI? I am not being facetious I genuinely don’t know how this could be constructed or implemented

Automated lint: 12 errors, 54839 warnings, 3 info

DeepSeek V4.1 Flash is now available on Together AI’s model hub (https://www.together.ai/models/deepseek-v4-1-flash). The release includes the DeepSeek V4.1 Flash API with updated pricing, performance

Wild 24 hours for AI and lots of different proposals have been made. TLDR; the only *tangible* new fact is that OpenAI and Anthropic are going to have embedded 3rd party evaluators from unknown organi

Terence Tao / What's new: A group of 25 Fields Medal recipients says AI companies' push to solve mathematical problems as a benchmark is detrimental to the science of mathematics — I am proud to be am

arXiv:2609.09875v1 Announce Type: new Abstract: Existing evaluation frameworks mostly assess only one part of AI agents, such as task completion (AgentBench) or security robustness (AgentDojo, ASB), r

arXiv:2609.11870v1 Announce Type: new Abstract: A language model normally begins training with random word embeddings: whatever 'banana' means must be learned from training corpora. I implement St. Au

Datasette 1.0a39 and 0.65.4 security releases Today we're releasing two new security patch versions of Datasette: 1.0a39 and 0.65.4 - one for the current alpha series and one for the stable 0.65.x fam

arXiv:2609.10421v1 Announce Type: cross Abstract: Background: Emergency Department (ED) return visits are commonly reviewed for quality assurance, but are often limited (e.g., to revisits within 48-72

arXiv:2603.17094v2 Announce Type: replace Abstract: Simulating human conversations using large language models (LLMs) has emerged as a scalable methodology for modeling human social interaction. This

arXiv:2603.15617v2 Announce Type: replace Abstract: Can AI make progress on important, unsolved mathematical problems? Large language models are now capable of sophisticated mathematical and scientifi

arXiv:2609.08861v1 Announce Type: new Abstract: Benchmark scores are a central currency in model releases: they inform purchasing decisions, shape public trust, and influence policy. Yet, a key assump

arXiv:2609.10144v1 Announce Type: cross Abstract: AI systems become more useful when they can adapt to the people using them, but in multi-agent systems, useful context learned by one agent often rema

arXiv:2609.10016v1 Announce Type: cross Abstract: We introduce MetroLLM-Bench, a 955-case benchmark for testing language models as the policy layer of a transit kiosk. It covers six real metro systems

This model is insane at landing pages. I asked DeepSeek V4.1 Flash & Claude Fable 5 to build me a landing page for a movie theater. Fable cost $1.21 while V4.1 Flash cost 2.6 cents, making it more tha

arXiv:2609.09949v1 Announce Type: cross Abstract: Real-world detectors must often interpret functional or ambiguous prompts, yet conventional models such as YOLO remain restricted to fixed class lists

arXiv:2609.08279v1 Announce Type: cross Abstract: Agent memory systems must discard stored information when their history exceeds a fixed token budget. Existing budget-accuracy frontiers quantify the

arXiv:2609.05441v1 Announce Type: new Abstract: Long-term memory for LLM agents is evaluated today by conversational recall benchmarks (LoCoMo, LongMemEval), which measure question answering over dial

Astra is awesome, but if you're using it specifically to parse and extract a massive volume of documents, you will break the bank (10c+ per page on parsing and extraction!) We've added a native LlamaP

arXiv:2609.06341v1 Announce Type: new Abstract: Linear algebra provides the framework of concepts (matrix rank, singular value decomposition (SVD), and eigendecomposition) that modern artificial intel

Automated lint: 12 errors, 53990 warnings, 3 info

arXiv:2609.05911v1 Announce Type: cross Abstract: Long-horizon LLM agents interact with untrusted content, persistent memory, external state, and sensitive tools. Existing analyses often characterize

arXiv:2609.08040v1 Announce Type: cross Abstract: The software supply chain has become an increasingly exposed attack surface because of its reliance on intricate yet fragile dependencies. Existing de

What is Astra actually supposed to be better at, Because if the answer is “everything,” then I genuinely do not understand what kind of testing was done for writing-heavy use cases. My biggest issue i

Tae Kim / Key Context: Astra working with Blender via computer use feels like magic, showing computer use could be the fourth demand wave after chatbots, reasoning, and agentic coding — AI computer to

arXiv:2609.04895v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models activate only a small subset of experts per token, but the full expert set often exceeds GPU memory, causing repeated we

I've built and open-sourced ParaRNN, a PyTorch/Triton library that parallelizes the training of non-linear recurrent architectures across time using Newton-Raphson iterations mapped to associative pre

Really strong benchmark paper on coding agents. Claude Opus 5 running under Claude Code passes 23.9% of the evaluations. An expert human reference scores 82.2%. Here is what the task actually is. The

arXiv:2609.05232v1 Announce Type: new Abstract: Autonomous AI agents increasingly select actions in environments whose memory, execution-time, runtime, compute, and operational constraints determine w

arXiv:2609.04453v1 Announce Type: cross Abstract: Expert pruning reduces the memory and serving cost of Mixture-of-Experts (MoE) models by removing low-importance experts identified by the router, ass