All entries

Gotta Catch them all: the modes of Sycophancy

arXiv:2607.20146v1 Announce Type: new Abstract: Large language models often align with users' beliefs at the expense of factual accuracy, a behavior known as sycophancy. Prior mechanistic studies larg

researcharxiv-cs-cl

GPT-5.5 Scores 10.6% on ActiveVision, Humans Hit 96.1% [R]

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

The interesting finding from a new [arXiv paper](https://arxiv.org/abs/2607.16165) isn't that a frontier vision model failed a new benchmark, that happens weekly, but the specific shape of the failure

model-releasesr-machinelearning

23 Jul 2026

Tutorials

GraphContainer: A Unified Platform for Comparing and Debugging Graph RAG Methods

arXiv:2607.19362v1 Announce Type: new Abstract: Graph RAG mitigates hallucinations and stale knowledge in LLMs, particularly for multi-hop question answering. However, existing approaches remain highl

tutorialsarxiv-cs-ai

Great paper on self-improving agent harnesses. (bookmark it) If you maintain a production agent harness, finding every file behind one behav…

Great paper on self-improving agent harnesses. (bookmark it) If you maintain a production agent harness, finding every file behind one behavior is often harder than writing the edit. Harness Handbook

model-releasesdair-ai--x

Great X: A Unified Multi-Modal Simulator Bridging the Sim2Real Gap for 6G

arXiv:2507.08716v4 Announce Type: replace Abstract: Large-scale, precisely synchronized multi-modal datasets are critical for data-driven sixth-generation (6G) wireless research, yet real-world collec

safetyarxiv-cs-cv

Group-of-Latents: Perceptual Video Compression at Extreme Bitrates via Masked Latent Generative Modeling

arXiv:2607.19437v1 Announce Type: cross Abstract: Most existing video compression algorithms follow a paradigm of transformation and quantization, optimizing the trade-off between distortion and bitra

researcharxiv-cs-cv

Applications

GTM strategy of every company in Forbes AI 50: → 90% use enterprise field sales → 46% combine enterprise sales + PLG → 44% use usage-metered…

applicationsyohei-nakajima--x

GTM strategy of every company in Forbes AI 50: → 90% use enterprise field sales → 46% combine enterprise sales + PLG → 44% use usage-metered pricing → The average startup runs 2.7 GTM motions → Conten

Guardrails as Scapegoats: Auditing Unfaithful Safety Refusals in Tool-Augmented LLM Agents

arXiv:2607.19449v1 Announce Type: cross Abstract: Evaluation frameworks for tool-augmented LLM agents focus overwhelmingly on capability metrics or explicit tool crashes, leaving silent infrastructure

safetyarxiv-cs-ai

H^2SD: Hybrid Hindsight Self-Distillation

arXiv:2607.18955v2 Announce Type: replace-cross Abstract: Reinforcement learning with verifiable rewards (RLVR) provides reliable outcome supervision for language model reasoning, but a scalar traject

researcharxiv-cs-cl

HalluTruthQA: A Fine-Grained Benchmark for Hallucination Detection, Localization, and Explanation in Arabic Question Answering

model-releasesarxiv-cs-cl

arXiv:2607.20219v1 Announce Type: new Abstract: Large language models (LLMs) can generate fluent Arabic answers, yet factual errors remain difficult to detect, localize, explain, and verify. Existing

Hard Guarantees at a Measured Price: Entropy-Stable Learned Finite Volumes for Compressible Flow

arXiv:2607.20171v1 Announce Type: cross Abstract: Learned solvers for compressible flow are usually compared to classical methods at equal mesh resolution rather than at equal computational cost, and

Harnessing Disagreement: Detecting Correlated Agreement Blindness in Multi-Agent Triage

arXiv:2607.19899v1 Announce Type: cross Abstract: Disagreement-triggered escalation can create a structural blind spot in multi-agent arbitration: as base learners improve, they tend to converge, weak

safetyarxiv-cs-lg

Tools

@HarryStebbings @lqiao Spotify https://open.spotify.com/episode/14rh372tSdEzRITBQz9HSP?si=c9ed4a9eae9f4de0 Youtube https://youtu.be/PCAiqKCf…

@HarryStebbings @lqiao Spotify https://open.spotify.com/episode/14rh372tSdEzRITBQz9HSP?si=c9ed4a9eae9f4de0 Youtube https://youtu.be/PCAiqKCfRSk?si=WDP2PfIkdn0XYH9T Apple Podcasts https://podcasts.appl

toolsfireworks-ai--x

Hazard or Anomaly? Evaluating VLMs for Understanding Dangers and Discrepancies

arXiv:2607.18325v1 Announce Type: new Abstract: Modern safety-critical systems increasingly rely on human-robot interaction to reduce disaster risk and support decision-making during emergencies. Visi

safetyarxiv-cs-cv

HeadCast: Casting Attention Heads for Efficient Autoregressive Video Generation

arXiv:2607.20125v1 Announce Type: cross Abstract: Autoregressive (AR) video diffusion models have become a promising paradigm for long and streaming video synthesis, but the continuously growing Key-V

Health in ChatGPT is starting to roll out to U.S. users. You can securely connect Apple Health and supported medical records to understand y…

Health in ChatGPT is starting to roll out to U.S. users. You can securely connect Apple Health and supported medical records to understand your information in context, track what has changed, and have

model-releasesopenai--x

Agents

High-risk autonomous behaviours are an increasingly prevalent and dangerous reality for frontier AI models https://www.wired.com/story/opena…

Frontier AI models are increasingly demonstrating high‑risk autonomous behaviours that pose safety threats. Incidents such as OpenAI‑released models escaping containment safeguards and a Hugging Face

agentsyoshua-bengio--x

HijackKV: New Threat in Position-Independent KV Cache Reuse

arXiv:2607.19957v1 Announce Type: cross Abstract: Key-Value (KV) cache reduces inference latency in large language models (LLMs). Traditional prefix-based reuse has low cache hit rates across inferenc

researcharxiv-cs-ai

HOMIE: Human-object Centric Video Personalization via Multimodal Intelligent Enhancement

arXiv:2607.18217v2 Announce Type: replace Abstract: Human-object centric video personalization (HOCVP) is a core task within subject-driven video generation. However, existing methods suffer from two

safetyarxiv-cs-cv

How Does Urban Context Relate to Residential Building Health? A Vision-POI Fusion Framework for Building-Level Housing Inspection

arXiv:2607.20263v1 Announce Type: new Abstract: Housing-level urban physical examination is essential for identifying residential building problems and supporting targeted urban renewal. Existing auto

researcharxiv-cs-cv

How Fast Can Reward Models Score? A Systems Study of C++ and PyTorch Inference Runtimes for RLHF

arXiv:2607.19712v1 Announce Type: new Abstract: In RLHF pipelines, reward scoring blocks policy updates. Slow scoring bottlenecks the entire loop, since no update runs until every rollout gets a score

safetyarxiv-cs-lg

Industry

Hugging Face might be one of the only companies making sure open-source AI wins and that people don’t end up as a permanent AI underclass

Hugging Face might be one of the only companies making sure open-source AI wins and that people don’t end up as a permanent AI underclass Introducing: The Stack v3 One thing that became very clear ove

industryclem-delangue--x

Hybrid LLM-Guided Search for Quantum Reservoir Architecture Design

model-releasesarxiv-cs-ai

arXiv:2607.19506v1 Announce Type: cross Abstract: Quantum reservoir computing (QRC) uses fixed quantum dynamics as a high-dimensional temporal feature map and trains only a lightweight classical reado

Hybrid LSTM-Graph Neural Framework for Robust Financial Fraud Detection and Adversarial Resilience

arXiv:2607.19350v1 Announce Type: new Abstract: Financial institutions face significant challenges in detecting sophisticated money laundering patterns, such as smurfing and layering, due to extreme d

researcharxiv-cs-ai

HyGRL: Adaptive Hybrid Graph Reasoning for Multi-Entity Questions

arXiv:2607.19398v1 Announce Type: new Abstract: Multi-entity compositional questions pose significant challenges to existing retrieval-augmented language models. Conventional methods fall into a dilem

safetyarxiv-cs-ai

HypEMBER: Hypernetwork-based Ensemble for Robust Policy Learning of Parametrized Dynamical Systems

model-releasesarxiv-cs-lg

arXiv:2607.19628v1 Announce Type: new Abstract: In this work we investigate reinforcement learning (RL) as a framework for the robust control of parametrized dynamical systems in presence of measureme

Hypothesis-and-Refinement Learning of Organic Structures from Multimodal Spectroscopic Data

model-releasesarxiv-cs-lg

arXiv:2607.19816v1 Announce Type: cross Abstract: Determining molecular structures from spectroscopic data remains fundamentally challenging because the inverse problem is intrinsically underdetermine

I built an open-source RAG chatbot starter that runs fully locally with Ollama (FastAPI + ChromaDB)

I kept re-wiring the same RAG plumbing on every project, so I turned it into a clean starter and open-sourced it. Upload a PDF, ask questions, and get answers with page-level source citations. It runs

model-releasesr-ollama

Industry

I listened to 85 minutes of The Economist’s interview of Elon so you don’t have to. Besides, it’s behind a paywall. Elon’s predictions: In f…

I listened to 85 minutes of The Economist’s interview of Elon so you don’t have to. Besides, it’s behind a paywall. Elon’s predictions: In five years, AI compute will exceed the sum of all human intel

industryelon-musk--x

I Made a Local Huggingface On My NAS

https://preview.redd.it/u8alj38wr0fh1.png?width=1860&format=png&auto=webp&s=3578776c60d9548a135a018702f44d0fddedd4b0 Little side project I'm doing so I can easily transfer any model I want fast to my

local-air-localllama

I run GLM-4.5-Air (110B) on 16Gb ram consumer machine and Qwen3-30B at 20 tok/s

In the past few months I’ve experimenting heavily and tortured my old 2016 Desktop PC to run the biggest Local LLM I can fit. I documented the whole process and research and I’ve published a repositor

local-air-ollama

I support open-source models distilling what commercial companies distilled for free from the entire internet. I published distillation for …

I support open-source models distilling what commercial companies distilled for free from the entire internet. I published distillation for free in 1991 in Europe - this was copied in the US and in Ch

researchdavid-ha--x

I trained a 0.5M model on 1B tokens of Fineweb-edu dataset.

model-releasesr-localllama

Hi everyone, About a month ago I publish my very first research paper on my neural network architecture called Silia. You can look at the model here: https://huggingface.co/Srijan-Srivastava/Silia-v2

Agents

I wrote the latest of my occasional guides to which AI to use right now for non-experts who want to get stuff done. The agentic systems avai…

I wrote the latest of my occasional guides to which AI to use right now for non-experts who want to get stuff done. The agentic systems available to everyone are getting extremely powerful (even as th

agentsethan-mollick--x

IBoxCLA: Towards Robust Box-supervised Segmentation of Polyp via Improved Box-dice and Contrastive Latent-anchors

model-releasesarxiv-cs-cv

arXiv:2310.07248v5 Announce Type: replace Abstract: Box-supervised polyp segmentation attracts increasing attention for its cost-effective potential. Existing solutions often rely on learning-free met

IConE: Batch Independent Collapse Prevention for Self-Supervised Representation Learning

arXiv:2603.15263v2 Announce Type: replace-cross Abstract: Self-supervised learning (SSL) has revolutionized representation learning, with Joint-Embedding Architectures (JEAs) emerging as an effective

Identity-Paired Progressive Depth Training: When Trainability Persists Beyond Expressibility

arXiv:2607.16800v2 Announce Type: replace-cross Abstract: Variational Quantum Algorithms (VQAs) are a leading paradigm for near-term quantum computing, yet their training suffers from sensitivity to c

If you are building real-time voice agents with @GoogleDeepMind Gemini Live, you can now trace your speech-to-speech agent loops directly in…

model-releasesharrison-chase--x

If you are building real-time voice agents with @GoogleDeepMind Gemini Live, you can now trace your speech-to-speech agent loops directly in @LangChain! - Speaker callback hooks capture only the exact

Applications

IGGT4D: Streaming 4D Instance-Grounded Geometry Transformer

arXiv:2607.19228v1 Announce Type: new Abstract: Real-world spatial intelligence requires agents to understand scenes from continuous video streams, where objects move, persist, disappear, and reappear

applicationsarxiv-cs-cv

Image Editing Models are Numerical Solvers

arXiv:2607.18787v1 Announce Type: new Abstract: We investigate whether a pretrained generative image-editing model can provide a common interface for numerical simulation. Physical inputs and solution

researcharxiv-cs-cv

IMMoE: Incomplete Multi-View Anomaly Detection via Mixture of View Experts Fusion

arXiv:2607.19032v1 Announce Type: new Abstract: Existing Multi-view Anomaly Detection (MAD) methods assume that all views are completely available and model each view separately. However, in real indu

local-aiarxiv-cs-cv

Importance-Aware OBS Pruning for Diffusion Models

model-releasesarxiv-cs-cv

arXiv:2607.20048v1 Announce Type: new Abstract: We propose importance-aware pruning for diffusion models, a training-free framework that prioritizes preserving parameters critical to semantically sali

In a leaked four-hour investor talk, DeepSeek's Liang Wenfeng says the main US-China gap is compute access, Nvidia's CUDA moat is disintegrating, and more (Fred Gao/Inside China)

Fred Gao / Inside China: In a leaked four-hour investor talk, DeepSeek's Liang Wenfeng says the main US-China gap is compute access, Nvidia's CUDA moat is disintegrating, and more — In a rare four-hou

model-releasestechmeme

In-Context Learning for Wound Classification with Small Multimodal Language Models

model-releasesarxiv-cs-cv

arXiv:2607.18819v1 Announce Type: new Abstract: Wound image classification is often treated as a task-specific supervised learning problem, requiring substantial amounts of manually labelled data and

In-Run Data Shapley for Adam Optimizer

arXiv:2602.00329v4 Announce Type: replace-cross Abstract: Reliable data attribution is essential for mitigating bias and reducing computational waste in modern machine learning, with the Shapley value

safetyarxiv-cs-ai

In-the-Flow Agentic System Optimization for Effective Planning and Tool Use

arXiv:2510.05592v2 Announce Type: replace Abstract: Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a singl

local-aiarxiv-cs-ai

inclusionAI/LLaDA2.2-flash · Hugging Face

model-releasesr-localllama

LLaDA2.2-flash is an agent-oriented diffusion language model in the LLaDA2 series. By introducing Levenshtein Editing (with DELETE and INSERT control tokens) to diffusion language modeling, it represe

Information Discernment in Large Language Models

model-releasesarxiv-cs-ai

arXiv:2607.19355v1 Announce Type: new Abstract: LLMs are increasingly used with external knowledge sources like the internet. Do they weigh information appropriately -- updating more for reliable sour