Search: “claude”

LLM Agents Can Easily Tamper With Their Own Traces

arXiv:2609.30266v1 Announce Type: cross Abstract: Asynchronous monitoring, incident investigations, and compliance audits primarily rely on agent traces to reconstruct what happened. These analyses as

Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMs

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

DGX agent

arXiv:2609.29775v1 Announce Type: cross Abstract: Large Language Models (LLMs) consume and produce a single sequence of text; hence, if text can be added to the beginning of the LLM's response, i.e.,

model-releasesarxiv-cs-ai

25 Sept 2026

Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models

arXiv:2609.30048v1 Announce Type: new Abstract: If a language model can recognize code it wrote, it may favor that code as a judge, and instances of one model monitoring each other could collude. We t

The Domestic Unprotected Zone: Algorithmic Governance and the Reproduction of Perpetrator Discourse in Conversational AI

arXiv:2609.28479v1 Announce Type: cross Abstract: Conversational AI increasingly mediates intimate-partner communication, and refusal logic at the inference layer now functions as a governance thresho

Watched my LLM router burn $11 on a 3-word prompt. Here's the fix.

I run a personal always-on agent fleet — eight seats claiming work in parallel from a shared task board, routing across Anthropic, OpenAI, xAI, Google, and a local Ollama floor. This week the spend fo

model-releasesr-ollama

After months of heads-down grinding, I’m proud to share that the @GPTZeroAI team has released the world’s most accurate AI detection model, …

model-releasesemad-mostaque--x

After months of heads-down grinding, I’m proud to share that the @GPTZeroAI team has released the world’s most accurate AI detection model, GPTZero 4o. It’s called 4o because there are four zeros: we

Ajar: Measuring Open Privilege in Agent Defenses

arXiv:2609.26900v1 Announce Type: cross Abstract: A language model agent acts through the tools it is given. The data it reads while working on a task can redirect what it does with those tools. A gro

FDE-Bench: Evaluating LLM Agents for Deployment Environment Configuration

arXiv:2609.27571v1 Announce Type: cross Abstract: Deployment requires an agent to turn application code into a running system whose services connect, become ready, and remain observable. FDE-Bench eva

Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents

arXiv:2609.27353v1 Announce Type: new Abstract: Web agents are usually evaluated in live environments, where environment state and judge models drift between runs, so the same checkpoint rarely reprod

I built a desktop benchmark for local Ollama models — TTFT, tok/s, VRAM, with a HuggingFace GGUF catalogue

Built this because I got tired of guessing whether a model would fit in my 6GB VRAM before downloading it, and because ollama.com's library doesn't let you sort by size. What it does: - Benchmarks Oll

model-releasesr-ollama

Interesting new paper from Google DeepMind. Studies agents that follow bad advice from users. Users often suggest a fix that sounds right an…

Interesting new paper from Google DeepMind. Studies agents that follow bad advice from users. Users often suggest a fix that sounds right and is wrong. XYEval adds one confident, misleading hint to ta

model-releasesdair-ai--x

Noob, Question.

I'm in a position where I didn't mean to go down this AI rabbit hole... I was looking into mental cognitions and how the mind is made up and learned that my old RTX 5k could run small models. So that

model-releasesr-ollama

Stravito integrates market research into enterprise AI tools with MCP server

model-releasessiliconangle

Knowledge management startup Stravito AB today introduced a Model Context Protocol server that lets people access the company’s consumer and market research from AI tools such as OpenAI Group PBC’s Ch

ThinkingCap 3.8-27B vs. Swift 3.8-27B vs. Qwen 3.8-27B Benchmarks

With the release of ThinkingCap-Qwen3.8-27B, I thought it would be worthwhile to do a comparison between the original Qwen3.8-27B, the new ThinkingCap, and Swift-Qwen3.8-27B. Both Swift which I alread

b11120

model-releasesllama-cpp-releases

vulkan: hide internal symbols to prevent duplicate-dlopen state destruction (#29139) Since #28732 our internal symbols are exported. A duplicate copy dlopened and dlclosed by ggml_backend_load_all() t

b11126

model-releasesllama-cpp-releases

vulkan: add IQ4_XS MMQ/MMV matmul kernels (#28415) vulkan: optimize IQ4_XS matmul kernels Assisted-by: OpenAI Codex vulkan: address IQ4_XS review nits drop the dead LOAD_VEC_A != 8 branch in the IQ4_X

CliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agents

arXiv:2609.26779v1 Announce Type: cross Abstract: Agents often work on complex problems that require millions of tokens of context, which necessitates compacting across sessions due to limited context

Et Tu, Brute? Economic Misalignment in Personal AI Agents

arXiv:2609.24927v1 Announce Type: new Abstract: Personal AI agents make recommendations and take actions on people's behalf in high-stakes economic contexts, e.g., buying a flight, choosing health ins

FinFIRST: Benchmarking Search Agents for Financial Information Retrieval, Sourcing and Traceability

arXiv:2609.25192v1 Announce Type: new Abstract: Financial search is a highly demanding task for LLM agents, requiring not only a correct final answer but also temporally valid information retrieval, a

Gemini 3.8 TTS Playground

model-releasessimon-willison

Tool: Gemini 3.8 TTS Playground Google released two new Gemini text-to-speech models today - gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts. They come with a library of over 2,000 voices, plus the

Local Ai

How SWE-Serve Exposes the Gap Between Local Tests and Live Serving

SWE‑Serve evaluates AI coding agents on 53 inference‑engineering tasks derived from SGLang pull requests across six functional domains. Among 19 tasks that launch a live server, patches that pass loca

local-ainvidia-developer

IntLawNER: A Named Entity Recognition Dataset and Benchmark in International Law

arXiv:2609.22529v1 Announce Type: new Abstract: International law provides the normative framework through which states coordinate action, regulate armed conflict, and protect human rights, yet its te

MiMo-V2.6-Pro is live on Wally. Day 1. @XiaomiMiMo's new open-weights flagship, top of the @ArtificialAnlys Intelligence Index for open-weig…

model-releasesyohei-nakajima--x

MiMo-V2.6-Pro is live on Wally. Day 1. @XiaomiMiMo's new open-weights flagship, top of the @ArtificialAnlys Intelligence Index for open-weight models. 320 tok/s, 2.4x the model maker's own API, in <24

Self-Cleaning and Captured Anyway: One Measured Primitive for Error in a Store an Agent Writes to Itself, and What a Falling Score Actually Measures

arXiv:2609.25052v1 Announce Type: new Abstract: 'An agent that writes its conclusions into a store it later retrieves from closes a loop usually reported as one-way contamination. Taking the loop to t

stuntd: a local Jev-compatible server on Laya that learns from your own traffic (no API key needed)

What it is. stuntd is a local proxy for typed LLM decisions: choice, yes/no, score. Your app keeps calling its provider through it. stuntd records the decisions, trains a small head on top of Laya (th

TelecomGPT-R1: Unified Post-Training for Reasoning Across Heterogeneous Telecom Tasks

arXiv:2609.25356v1 Announce Type: new Abstract: Large language models (LLMs) offer great potential to automate a broad range of telecom engineering tasks by reasoning over standards, network configura

The sad thing is that Dario knows better. He was a PhD student. He knows the significance level of this result. He knows that if he had walk…

model-releasesgary-marcus--x

The sad thing is that Dario knows better. He was a PhD student. He knows the significance level of this result. He knows that if he had walked into Bill’s office (his advisor) with “we found an intere

There are countless CRISPR-like sequences hidden in genomes of many organisms (eg @Basecamp_Res). Well-known. Vast majority not interesting.…

model-releasesgary-marcus--x

There are countless CRISPR-like sequences hidden in genomes of many organisms (eg @Basecamp_Res). Well-known. Vast majority not interesting. This isn’t a noteworthy discovery. The hype level leads me

this is a new type of post we're trying where we share in-depth about how we do work using specific prompts and techniques you should be abl…

this is a new type of post we're trying where we share in-depth about how we do work using specific prompts and techniques you should be able to replicate let us know if it's useful or if you have any

model-releasesthariq--x

Trains but Doesn't Learn: A Post-Training Delivery Benchmark for LLM Agents as Forward-Deployed Engineers

arXiv:2609.25237v1 Announce Type: cross Abstract: Post-training is becoming a service (PTaaS): a customer hands an operator data and a goal, and a forward-deployed engineer (FDE) returns a fine-tuned,

Auditing Political Alignment in LLM Assistants: Engagement, Stance, and User Identity

arXiv:2609.23039v1 Announce Type: new Abstract: LLM-based AI systems answer political questions for hundreds of millions of people. Current audits measure what they say to an average user, but their b

Beyond Accuracy and Surface Fluency: Risk-Sensitive Evaluation of LLMs for Legal Clause Generation

arXiv:2609.22127v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used to draft contractual language, yet conventional accuracy or preference-based evaluations are poorly m

Can Coding Agents Reproduce Official Statistics? Metadata, Retry Budget and the Limits of Execution Feedback in a Controlled Eurostat Benchmark

arXiv:2609.22222v1 Announce Type: cross Abstract: Large language models can generate executable data-analysis code, but successful execution is not equivalent to a valid official-statistics result. Th

Defusing Explosive Prompts: Understanding and Preventing Trigger-Based Prompt Injections in LLM Agents

arXiv:2609.22510v1 Announce Type: cross Abstract: As LLM applications integrate with external tools, they are increasingly exposed to indirect prompt injection (IPI), where adversarial instructions ar

Did a few quick tests and these seem very very legit. 2.6 Pro in particular is doing well in some pretty hard tasks.

model-releasesclem-delangue--x

Did a few quick tests and these seem very very legit. 2.6 Pro in particular is doing well in some pretty hard tasks. Introducing Xiaomi MiMo-V2.6 — Pro & Flash. Frontier intelligence, all the modaliti

GameReplica: A Benchmark for Black-Box Visual Game Replication by Vision-Language Agents

model-releasesarxiv-cs-cv

arXiv:2609.22308v1 Announce Type: new Abstract: Coding-agent benchmarks usually evaluate implementation after the target behavior has been specified in text, code, or demonstrations. Existing research

GPT-6 Sol costs 2/1M input and 10/1M output tokens, and GPT-6 Luna costs 0.10/1M input and 0.50/1M output tokens, both about 50% cheaper than GPT-5.6 (Nat Rubio-Licht/The Deep View)

Nat Rubio-Licht / The Deep View: GPT-6 Sol costs 2/1M input and 10/1M output tokens, and GPT-6 Luna costs 0.10/1M input and 0.50/1M output tokens, both about 50% cheaper than GPT-5.6 — OpenAI is relea

model-releasestechmeme

MimicAgent: Quadruped Skills via Prompt-to-Trajectory Generation

model-releasesarxiv-cs-ro

arXiv:2609.24145v1 Announce Type: new Abstract: We present MimicAgent, a prompt-to-trajectory generation framework for learning dynamic quadruped skills. Although reward shaping is extensively used wh

MobileCybench: Evaluating Agent Vulnerability Discovery via Executable Probes

arXiv:2609.23980v1 Announce Type: cross Abstract: AI agents now report vulnerabilities faster than maintainers can review them. Reports often depend on security properties specific to the application,

OSWorld-Pro: Process-based Evaluation for Computer Use Agents

arXiv:2609.24890v1 Announce Type: new Abstract: Evaluation of Computer-Use Agents (CUAs) is often limited to the final deliverables they create (at the end of hundreds of steps) and assessed with func

The Situated Identity Test: Distinguishing Persistent Cognitive Identity from Persona Imitation

arXiv:2609.22195v1 Announce Type: new Abstract: Large language models can convincingly adopt personas, recall past dialogues, and weave rich autobiographies. Yet this conversational eloquence conceals

Triggers and Diagnostics for LLM-Based Interpretability Failures in Active Inference Agents

arXiv:2609.23215v1 Announce Type: new Abstract: LLM explainers are increasingly attached to autonomous agents as runtime oversight, with operators reading a generated account of the agent's beliefs an

BREAKING: Grok 4.7 dominates the Legal Agent Benchmark. It scored 19.6% on realistic, long-term legal tasks, beating every other model teste…

model-releaseselon-musk--x

BREAKING: Grok 4.7 dominates the Legal Agent Benchmark. It scored 19.6% on realistic, long-term legal tasks, beating every other model tested. That is nearly 3× Fable 5.1. Another major win for Grok 4

GameLogicBench: Evaluating Coding Agents on Runtime Game Logic with Tick-Level State Assertions

arXiv:2609.21562v1 Announce Type: cross Abstract: Coding agents can modify and test code across large software projects. Game development is a domain where agents must implement gameplay rules. A game

I am reminded of the fact that the infamous Milgram Experiment is actually quite weak because most people suspected the shocks were fake (ki…

model-releasesethan-mollick--x

I am reminded of the fact that the infamous Milgram Experiment is actually quite weak because most people suspected the shocks were fake (kind of obvious in retrospect) & thus went along with shocking

mini-AGI - dynamically grown (530M params currently and growing) continual learning model trained from scratch on 8GB VRAM laptop from batch-1 stream of data.

Sorry for the pretentious name, I know, I know.. It just contains all the pieces I would like to see a AGI model to have, and I can't stand the temptation. Before throwing rocks at me, please take a g

[Splash Engine] Qwen3.8-27B in native 8-bit at 37–55 tok/s on Apple Silicon: Extending Splash to Q8, 256k context scaling, and the 'Reasoning Cliff'

https://preview.redd.it/nulsv53o8vqh1.png?width=4500&format=png&auto=webp&s=74765dbd409f4c221640f9f6000a685f6fdbb242 Spent weekend benchmarking the Splash engine (by Incoai) and extending its architec

MCP was always a bad idea?