Search: “models”

Decoupling semantics from vision: A framework for faithful visual-text compression evaluation

arXiv:2608.01848v1 Announce Type: new Abstract: Recent visual-text compression (VTC) methods, typified by DeepSeek-OCR, report impressive high token compression ratios for long-context modeling tasks

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

DeepSeek v4 Flash vs. Qwen3.6-27B, 3.5-122B, and Gemma 4 31B Benchmark

DGX agent

Just wanted to share my agentic coding benchmark run of DSv4F 0731 at both High and Low reasoning efforts (not Max)... I ran a 109-question subset of Aider Polyglot (the JS/C++/Python languages), base

model-releasesr-localllama

4 Aug 2026

Defending Membership Inference Attacks via Privacy-aware Sparsity Tuning

arXiv:2410.06814v2 Announce Type: replace Abstract: Over-parameterized models are typically vulnerable to membership inference attacks, which aim to determine whether a specific sample is included in

DeltaFlow: Noise-Adaptive Bidirectional Gated Delta Networks for Embedded Language Flows

arXiv:2608.01240v1 Announce Type: new Abstract: Embedded Language Flows (ELF) rely primarily on full non-causal attention for iterative denoising, repeatedly incurring quadratic sequence-mixing cost a

ET-Prune: Evidence-Aware Dynamic Budgeting for Visual Token Pruning in Text-Rich MLLMs

arXiv:2608.01979v1 Announce Type: cross Abstract: Visual token pruning reduces the inference cost of multimodal large language models, but a fixed token ratio is poorly matched to text-rich inputs. In

Feed-Forward Steering in Transformer Residual Dynamics

arXiv:2608.02071v1 Announce Type: new Abstract: Attention-only dynamical theories model Transformer residual directions as particles aggregating on a sphere. We extend this framework by incorporating

GPT-OSS has turned one year old today!

It is one of the best local models ever released, in both 20B and 120B versions. I always come back to it, especially the 120B version. Its only competition is, in my opinion, Qwen 3.5 122B, but that

Grounding Agentic VLMs with Dedicated Segmentation for Fine-Grained Vehicle Damage Assessment

arXiv:2608.02470v1 Announce Type: new Abstract: Vision-language models (VLMs) are increasingly deployed as reasoning agents in real-world visual assessment pipelines, yet their spatial grounding remai

HarMoE: Multi-Source Chest Radiograph Pretraining with Dataset-Disentangled Experts

arXiv:2608.02252v1 Announce Type: new Abstract: Recent vision-language models for chest X-ray understanding are largely built on image-report alignment and therefore rely heavily on MIMIC-CXR as the d

Humans Are More Diverse: Frontier LLMs Show Extreme Policies in Idealised AI Development Races

arXiv:2608.01193v1 Announce Type: cross Abstract: An AI development race creates a multi-agent safety dilemma. Each company can develop slowly and safely, or move faster while taking a risk that may r

InteracVid: Building a Real Interactive Audio-Visual Response Dataset from Live-Chat Videos

arXiv:2608.01157v1 Announce Type: new Abstract: Large language models have made text the default medium for human--AI interaction, buttext alone cannot express the full range of responses required by

Research

KING: Embodiment-Aware Kinematic Graph Neural Network for Unified Motion Representation of Legged and Wheeled Robots

arXiv:2608.01015v1 Announce Type: new Abstract: Kinematic models provide reliable motion constraints for odometry estimation in featureless environments, where exteroceptive sensing degrades and IMU i

researcharxiv-cs-ro

LangFIR: Discovering Sparse Language-Specific Features from Monolingual Data for Language Steering

arXiv:2604.03532v2 Announce Type: replace Abstract: Large language models (LLMs) show strong multilingual capabilities, yet reliably controlling the language of their outputs remains difficult. Repres

Local Ai

Language Equality has a Price: A Systematic Investigation of Multi-turn LLM Performance for EU-24+

arXiv:2608.01395v1 Announce Type: new Abstract: We evaluate large language models (LLMs) as language agents playing goal-directed dialogue games in self-play across 30 languages: the 24 official EU la

local-aiarxiv-cs-cl

llm-anthropic 0.26

model-releasessimon-willison

Release: llm-anthropic 0.26 Includes new features enabled by LLM 0.32: New models: claude-fable-5, claude-sonnet-5, and claude-opus-5. #75, #76 Added server-side tools for WebSearch, WebFetch, CodeExe

LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks

arXiv:2608.01964v1 Announce Type: new Abstract: Large language model (LLM) agents increasingly undertake long-horizon tasks that require sustained reasoning, tool use, and revision across many interde

Machine-Precision Prediction of Low-Dimensional Chaotic Systems from Noise-Free Data

arXiv:2507.09652v2 Announce Type: replace-cross Abstract: Low-dimensional chaotic systems such as the Lorenz-63 model are commonly used to benchmark system-agnostic methods for learning dynamics from

onepot-Bench 0: towards lab-aware in silico chemistry benchmarks

arXiv:2608.02595v1 Announce Type: new Abstract: Language models are playing an increasingly important role in laboratory science, performing tasks such as experiment planning, execution, and post-hoc

OpenART: Scaling Agent Red Teaming via Open-Ended Environment Evolution

arXiv:2608.00677v1 Announce Type: new Abstract: AI agents operate in persistent environments where early state changes can influence decisions far into the future. Unlike conventional language-model i

Passing Coarse Marginal Checks Can Be Cheap: Persona Mixtures and Imprecise Treatment-Response Estimates in an LLM Persona Panel

arXiv:2608.00979v1 Announce Type: cross Abstract: Large language models are increasingly used as synthetic research participants and are often validated by whether their marginal responses resemble hu

Research

PhyCheck: Fine-Grained Evidence-Grounded Dataset for Physical Law Understanding in Video-LLMs

arXiv:2608.02150v1 Announce Type: new Abstract: Embodied intelligence and world models require video understanding systems to go beyond recognizing objects and actions and develop an understanding of

researcharxiv-cs-cv

RAP: KV-Cache Compression via RoPE-Aligned Pruning

arXiv:2602.02599v4 Announce Type: replace Abstract: Long-context inference in large language models (LLMs) is bottlenecked by the memory and compute of the key-value (KV) cache. Structured pruning is

Remember OpenAI tripled its score on ARC-AGI with a harness fix? We fixed harness for Kimi K3 on CyberGym E2E: 2x better at vulnerability de…

model-releasesfireworks-ai--x

Remember OpenAI tripled its score on ARC-AGI with a harness fix? We fixed harness for Kimi K3 on CyberGym E2E: 2x better at vulnerability detection, 3x at patching K3 is SOTA cyber-defense model you c

S^4R: Selective Sampling, Subspaces, and Sparse Reconstruction for Compressed Long-Context KV Caching

arXiv:2608.00528v1 Announce Type: new Abstract: The growth of context window lengths in Large Language Models (LLMs) significantly enhances their long-context capabilities but incurs prohibitive memor

SCHEDBench: A Benchmark for Evaluating LLM Constraint Faithfulness in Natural-Language Combinatorial Scheduling

arXiv:2608.00991v1 Announce Type: cross Abstract: This paper introduces SCHEDBench, a natural-language benchmark for evaluating combinatorial scheduling constraint faithfulness under surface-form vari

SLMs as Multi-Agent Routers: A Progressive SFT and Reinforcement Learning Approach

arXiv:2608.00030v1 Announce Type: new Abstract: Specialised retrieval agents typically surface higher quality results than general-purpose search, but selecting the optimal agent for a given query rem

Test-Time Curriculum for Open-Set AIGC Detection

arXiv:2608.00559v1 Announce Type: new Abstract: AI-generated image detectors deployed in open-world environments inevitably face distribution shifts as new and stronger generative models continue to e

The UK AISI says it observed a total of 19 instances where Mythos and GPT-5.6 Sol tried to hack people and companies during a routine cyber evaluation in July (Sam Sabin/Axios)

Sam Sabin / Axios: The UK AISI says it observed a total of 19 instances where Mythos and GPT-5.6 Sol tried to hack people and companies during a routine cyber evaluation in July — The U.K. AI Security

model-releasestechmeme

UpliftBench: Revealing Outcome-Regime and Objective Mismatch in Uplift Evaluation

arXiv:2608.00915v1 Announce Type: new Abstract: Uplift modeling (conditional-average-treatment-effect estimation) drives personalized targeting, yet published uplift benchmarks frequently disagree on

Writing-System-Level Tokenizer Adaptation for Byte-Level BPE

arXiv:2608.00582v1 Announce Type: new Abstract: Pretrained byte-level BPE tokenizers can segment underrepresented languages inefficiently. Replacing a tokenizer changes the meaning of nearly every tok

Z-PEFT: Zero-shot Backdoor Detection in Parameter-Efficient Fine-Tuning via Canonical Spectral Signatures

arXiv:2608.02271v1 Announce Type: new Abstract: Parameter-Efficient Fine-tuned (PEFT) models are frequently downloaded from open repositories by practitioners. This widespread practice creates a signi

AI9Stars released G9v3-39A5B

AI9Stars has released G9v3-39A5B an open weights language model designed to deliver even stronger reasoning capabilities than ai9stars/G9v3-3B with its 39B and 5 active experts. It is released under t

Bootstrapping Self-Supervised Learning of Binary Classification Using Error Bounds: A Case Study on a Robotic Insertion Task

model-releasesarxiv-cs-ro

arXiv:2607.29640v1 Announce Type: new Abstract: Flexible manufacturing requires rapid deployment of solutions and minimal setup time to remain competitive. An essential attribute is the ability to con

Can LLMs Really Understand Item Difficulty Levels? Implications for Automated Item Generation Using LLMs

arXiv:2607.28634v1 Announce Type: new Abstract: The estimation of item difficulty plays a key role in both formative assessment and large-scale high-stakes summative assessments. This study explores h

Döner Bench DeepSeek-V4-Flash IQ2_XS running on a single RTX 3090

https://preview.redd.it/3zcvpbds14hh1.png?width=1911&format=png&auto=webp&s=a79aafb71eeca97638da93d2591902631e897fd5 I tried a test similar to the recent model-quant comparisons, but this time I focus

Frontier AI performance across the business disciplines: a case-grounded benchmark of knowledge work and analytical reasoning

arXiv:2607.16057v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are improving rapidly as reflected in benchmark scores, yet these AI benchmarks largely test capabilities such as

Hy-MultiTurn: A Six-Dimensional Benchmark for Deep Multi-Turn Dialogue Understanding

arXiv:2607.29196v1 Announce Type: new Abstract: Long-running multi-turn interactions with chatbots and agents are now common, and a correct response often depends on remembering earlier details, track

Safety

Incorporating data drift to perform survival analysis on credit risk

arXiv:2601.20533v2 Announce Type: replace-cross Abstract: Survival analysis has become a standard approach for modelling time to default by time-varying covariates in credit risk. Unlike most existing

safetyarxiv-cs-lg

Latent Sculpting for Zero-Shot Generalization: A Manifold Learning Approach to Out-of-Distribution Anomaly Detection

arXiv:2512.22179v3 Announce Type: replace Abstract: Detecting previously unseen attacks remains a major challenge for machine learning-based intrusion detection systems. Deep models trained on network

Locally Consistent Transductive Information Maximization for Few-Shot Remote Sensing Scene Classification

arXiv:2607.29192v1 Announce Type: new Abstract: Remote sensing scene classification is increasingly relying on foundation models pre-trained on large-scale Earth-observation data. Moreover, transducti

Looks Right, Works Right: A Project-Level Benchmark for Multi-Screen Mobile App Generation

arXiv:2607.28645v1 Announce Type: cross Abstract: Recent multimodal large language models can convert visual designs directly into executable code, but real mobile products require multiple screenshot

Research

Metaphor-Induced Algorithmic Steering: Cross-Domain Procedural Transfer in LLM Code Generation

arXiv:2607.28683v1 Announce Type: cross Abstract: Large language models benefit from elements in natural language, such as metaphors and analogies in training data and inference input to achieve gener

researcharxiv-cs-ai

MirrorCraft: Paired Evaluation under Hidden Rule Changes in Minecraft

arXiv:2607.29218v1 Announce Type: new Abstract: With the prosperity of the large language models (LLMs), it has become an interesting topic: how do LLM-based agents work in Minecraft? Unfortunately, m

MoPET: Parameter-Efficient Mixture-of-Experts for Unified Medical Image Classification

arXiv:2607.29462v1 Announce Type: cross Abstract: Adapting deep learning models to profound clinical heterogeneity typically relies on parameter-efficient fine-tuning (PEFT) to avoid the severe overfi

Safety

Persistent Convolution: A Topological Framework for AI Alignment Testing and Semantic Space Characterization

arXiv:2607.29008v1 Announce Type: cross Abstract: Modern opaque AI models prize performance over interpretability, which makes testing difficult. However, formal statistical tests conducted on a model

safetyarxiv-cs-lg

Question about Quant versus Size.

Sorry if this is asked a lot, but I was wondering if there is any clear winner on the Quantization versus Model Size debate? I can run Qwen3.6 27b at Q8, Laguna at Q6, and the new Deepseek Flash at Q3

Safety, or Just Capability? A Validity Audit of Agent-Safety Benchmarks

arXiv:2607.28685v1 Announce Type: new Abstract: Agent-safety benchmarks measure different behaviors, and their scores get quoted interchangeably as an agent's safety. We treat four of them (R-Judge, I

SEDR-Seq2P: A Lightweight Dilated Residual Sequence-to-Point Network for Multi-Task Industrial NILM

arXiv:2607.28693v1 Announce Type: cross Abstract: Industrial NILM remains challenging because measurement noise and widespread concurrent machine operation reduce the generalization of models tuned on