Search: “reasoning”

Annotations as Rollouts: Efficient and Scalable Reinforcement Learning for Video MLLMs

model-releasesarxiv-cs-cv

arXiv:2608.20492v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) have become a prevailing paradigm for unified video perception. However, post-training on large multi-task data

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

[Benchmark] Optimal DFlash2 quants for speed and context size, 5090 RTX, llama.cpp, Qwen 3.8 27B Dynamic3 Unsloth. Comparison with MTP

DGX agent

Graph: A cumulative metric of Speed x Context Size - the higher the dot - the better. Helpful for looking for the most balanced solution. The research relies on DFlash2 Q2 work by AnalogAlok: https://

model-releasesr-localllama

24 Aug 2026

Bridging Language and Spherical Space: Object-Centric Control for Text-to-Panorama Generation

arXiv:2608.20691v1 Announce Type: new Abstract: Panoramic image generation is increasingly important for immersive applications such as virtual reality, augmented reality, and 3D content creation. Unl

safetyarxiv-cs-cv

Tutorials

CDRL: Certification-Driven Reinforcement Learning for Neutrino Flavor Model Discovery

arXiv:2608.20686v1 Announce Type: new Abstract: Many scientific discovery problems require searching combinatorial hypothesis spaces under complex domain constraints. Reinforcement learning (RL) offer

tutorialsarxiv-cs-ai

Tutorials

Decoupled Vision-Language System for Multimodal Understanding and Generation

arXiv:2608.20382v1 Announce Type: new Abstract: We introduce a new architecture design for multimodal large language models (MLLMs), Libra, capable of both multimodal understanding and generation. Lib

tutorialsarxiv-cs-cl

Delay-corrected Bellman operator + causal attribution for constrained RL contraction proof under unknown stochastic delay [R]

model-releasesr-machinelearning

Standard constrained RL assumes consequences are immediate and attributable to the current action. This breaks down whenever violations are delayed and stochastic, which is most real-world settings yo

Don't Solve, Just Compare: Tiny Advisors for Runtime Intervention in LLM Agents

arXiv:2608.21027v1 Announce Type: new Abstract: LLM agents are emerging as an important paradigm for real-world tasks that require reasoning, tool use, and sequential decision-making. As these agents

Edge-Based Agentic Retrieval-Augmented Generation for Autonomous FHWA Bridge Inspection Compliance

arXiv:2608.20372v1 Announce Type: cross Abstract: The Federal Highway Administration (FHWA) mandates that over 600,000 bridges in the United States be evaluated against the Recording and Coding Guide

Harmonic Torsional Diffusion for Protein-Ligand Flexible Docking

model-releasesarxiv-cs-lg

arXiv:2608.20366v1 Announce Type: cross Abstract: Molecular docking requires reasoning jointly about ligand pose and protein flexibility. Most diffusion-based docking models predict torsional updates

Is Multimodal Speculative Decoding Ready for Diffusion-Based Parallel Drafting? A Survey and Empirical Diagnosis

arXiv:2608.20743v1 Announce Type: new Abstract: Speculative decoding accelerates autoregressive generation by allowing a lightweight drafter to propose future tokens while a target model verifies them

safetyarxiv-cs-ai

Multilingual Verifier Bias in RLVR: Benchmark, Rollout Diagnosis, and the Cross-Lingual Selection Bottleneck

arXiv:2608.20362v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) is a standard recipe for training large language models on mathematical reasoning, where an answer

PhotoBench: Beyond Visual Matching Towards Personalized Intent-Driven Photo Retrieval

arXiv:2603.01493v2 Announce Type: replace-cross Abstract: Personal photo albums are not merely collections of static images but living, ecological archives defined by temporal continuity, social entan

Industry

Physical AI’s moment has arrived – but moving from demo to deployment is the hard part. AWS wants to fix that

Physical artificial intelligence is emerging as the next major phase of AI. These systems not only generate content or analyze data but also perceive, reason about and act in the physical world. The o

industrysiliconangle

When does fusing hand-crafted knowledge with learned representations pay? A cost-normalized benchmark of stacking, substitution, and interference

model-releasesarxiv-cs-cv

arXiv:2608.21098v1 Announce Type: new Abstract: Fusing prior knowledge with data-driven learning is attractive where data is scarce, yet no controlled account says when it helps, is redundant, or harm

When Failures Propagate: Causal Failure Attribution in Agentic Retrieval-Augmented Generation

arXiv:2608.20627v1 Announce Type: cross Abstract: Agentic retrieval-augmented generation (RAG) interleaves retrieval, reasoning, and answer generation across multiple hops. A retrieval error at hop 1

When Graph-JEPA Learns the Wrong Thing: Diagnosing and Repairing Category-Conditional Collapse

arXiv:2608.20516v1 Announce Type: new Abstract: Joint-embedding predictive architectures are selected almost universally by linear probing and effective rank. We report a case where both read healthil

researcharxiv-cs-lg

Benchmark results: what is the best and fastest engine to run Qwen3.8-27B on macOS

WARNING: I think I made a mistake with omlx, and its results below seem to be without any kind of acceleration. I am rerunning them with Lightning MTP and ANE support for prefill. Will update the post

23 Aug 2026

Why has DSv4 0813 become trash?

It performs at full capacity without any issues anywhere, but only the Ollama DeepSeek seems particularly trashy. Why on earth has the Ollama DeepSeek unusable trash that goes into over reasoning and

model-releasesr-ollama

23 Aug 2026

I benchmark DFlash 2 (PR build) in llama.cpp on Qwen 3.8 27B against all speculative methods for 3 days. 2.26x on 100 real coding prompts, 4.68x with one n-gram drafter on top. Up to 8x on specific cases.

Hey guys, Inco AI shipped DFlash 2 a few days ago with a drafter for Qwen 3.8 27B and a llama.cpp PR. I built the PR and ran it against plain decoding, MTP, the n-gram lookup drafters, and my July DFl

22 Aug 2026

One of the interesting properties we’ve observed around schema-guided, complex document extraction tasks 📑 is that coding agent harnesses a…

model-releasesjerry-liu--x

One of the interesting properties we’ve observed around schema-guided, complex document extraction tasks 📑 is that coding agent harnesses are good baselines (in terms of cost/accuracy) for longer docu

22 Aug 2026

Single RTX 5090: Qwen3.8-27B NVFP4 at a real 262K context in vLLM — 77 tok/s short-context, 64.7 tok/s at 128K

This is the Qwen3.8-27B setup I actually use every day on one RTX 5090. I wanted to write it down with enough detail that another 5090 owner can reproduce it instead of guessing which memory knobs I u

22 Aug 2026

Continuous Adversarial MeanFlow Transfer

arXiv:2608.19540v1 Announce Type: cross Abstract: Training fast generators on new domains with limited data remains challenging for two reasons. First, adapting a pretrained diffusion or flow model to

researcharxiv-cs-cv

DeepSeek-V4-Flash-Vision-Exp is now live on the DeepSeek API Platform! 🚀 🔹 This experimental multimodal model matches DeepSeek-V4-Flash on…

model-releasesdeepseek--x

DeepSeek-V4-Flash-Vision-Exp is now live on the DeepSeek API Platform! 🚀 🔹 This experimental multimodal model matches DeepSeek-V4-Flash on text capabilities—including agents, reasoning, and world know

Evidence-Gated Task and Motion Planning with Vision-Language Models

arXiv:2608.20084v1 Announce Type: cross Abstract: Robots executing long-horizon manipulation tasks from natural-language instructions must reason about both semantic task structure and geometric feasi

FormalTCS: Benchmarking End-to-End Frontier Formal Theoretical Computer Science Research of Large Language Models

arXiv:2608.20153v1 Announce Type: new Abstract: Large language models (LLMs) have shown growing potential for automated theoretical computer science (TCS) research, yet existing benchmarks remain far

I benchmarked Ox Alpha on SWE-bench Verified Mini (50 tasks): 96% resolved. Now I’m skeptical of myself.

TL;DR: The result looks TOO GOOD TO BE TRUE — that's exactly why I'm posting it. I ran Ox alpha on the entire SWE-bench Verified-Mini set with the official mini-swe-agent scaffold — the same agent use

Tutorials

Orthogonal JEPA: Factorized Predictive States for Latent World Models

arXiv:2608.20065v1 Announce Type: new Abstract: World models construct latent states that support prediction, planning, and reasoning about an underlying system. Joint-embedding predictive architectur

tutorialsarxiv-cs-lg

Reward-Guided Autoregressive Graph Generation for Efficient Multi-Agent Communication Topology Design

arXiv:2608.20099v1 Announce Type: cross Abstract: LLM-based Multi-Agent Systems (MAS) achieve strong performance on complex reasoning tasks by coordinating multiple agents, but at the cost of substant

agentsarxiv-cs-cl

ROBOSHACKLES: A Safety Dataset for Human-Injury Prevention in Embodied Foundation Models

model-releasesarxiv-cs-ro

arXiv:2606.18632v2 Announce Type: replace Abstract: Embodied Foundation Models (EFMs) integrate multimodal understanding, future-state reasoning, and executable robot actions. Yet their safety alignme

The Third Restructuring of Software Form: From the Three-Tier Architecture to Storage, Models, and Agents

arXiv:2608.20201v1 Announce Type: new Abstract: Software form has undergone two paradigm shifts since its inception: Software 1.0, in which instructions determine behavior, and Software 2.0, in which

When Contextual Inference Fails: Cancelability in Interactive Instruction Following

arXiv:2603.19997v2 Announce Type: replace Abstract: We investigate the separation of literal interpretation from contextual inference in a collaborative block-building tasks, where an agent must resol

Applications

When Machines Speak: A Unified Generative Framework for Integrating Machine-Native Symbols into Pretrained Large Language Models

arXiv:2608.19529v1 Announce Type: cross Abstract: Many real-world AI systems represent entities, behaviors, and structured information using discrete machine-native symbols rather than natural languag

applicationsarxiv-cs-ai

Your Turn: At Home Turning Angle Estimation for Parkinson's Disease Severity Assessment

arXiv:2408.08182v5 Announce Type: replace-cross Abstract: People with Parkinson's Disease (PD) often experience progressively worsening gait, including changes in how they turn around, as the disease

10 questions every startup should answer before moving to production with their AI prototype

model-releasesgoogle-cloud-ai

It’s never been easier to start an AI-powered startup on Google Cloud. You grab an API key from Google AI Studio at breakfast, paste it into Antigravity, and by lunch you’ll have a nascent prototype o

Allocating Recurrent Compute in Looped Language Models

arXiv:2608.18230v1 Announce Type: new Abstract: Looped language models improve reasoning and knowledge manipulation by applying shared computation repeatedly. Existing systems usually repeat an entire

researcharxiv-cs-lg

Beyond Placement and Articulation: Usage-Driven Code Scenes for Embodied Interaction

arXiv:2608.18840v1 Announce Type: cross Abstract: Indoor scene synthesis provides essential environments for embodied AI, robotic manipulation, and simulation-based policy learning. Recent code-based

safetyarxiv-cs-cv

Bidirectional representational alignment between biological and artificial neural networks

arXiv:2608.18244v1 Announce Type: cross Abstract: Recent work has shown that representational alignment between biological and artificial neural networks is asymmetric: model representations predict n

safetyarxiv-cs-ai

Breaking the weakest link to evade vision language models

arXiv:2608.18938v1 Announce Type: new Abstract: Vision Language Models (VLMs) have recently emerged as a critical component of multimodal AI systems, enabling joint reasoning over visual and textual i

Decomposing Wrong-Consensus Agreement in LLM Self-Consistency: A GPT-4.1 Case Study

arXiv:2608.18795v1 Announce Type: cross Abstract: Majority voting over multiple LLM samples is widely used to raise answer accuracy, yet its gain varies erratically: on hard questions it can even back

Different Facets of Verbalised Overconfidence: an Interpretability Study

arXiv:2608.18106v1 Announce Type: cross Abstract: Large language models tend to overconfidence, giving assertive answers when the evidence suggests hedging or abstention. Using controlled reasoning sc

Evaluating and Explaining Prompt Sensitivity of LLMs Using Interactions

arXiv:2608.18539v1 Announce Type: cross Abstract: The remarkable capabilities of large language models (LLMs) are often undermined by their instability. Even subtle and semantically irrelevant changes

Google is a Leader in the 2026 Gartner® Magic Quadrant™ for Cloud-Native Application Platforms

model-releasesgoogle-cloud-ai

We are thrilled to announce that Google has been recognized as a Leader for the third year in a row in the 2026 Gartner® Magic Quadrant™ for Cloud-Native Application Platforms (CNAP). We believe this

Looped Language Models Improve Compositional Tool Calling

arXiv:2608.18171v1 Announce Type: new Abstract: Looped language models have shown promising results on reasoning benchmarks, yet their potential for agentic tool use remains largely unexplored. We stu

LT-Mem: Volatility-Aware Spatio-Temporal Memory for Lifelong Scene Understanding

arXiv:2608.19059v1 Announce Type: cross Abstract: Long-term robot operation in evolving environments requires object-level understanding that persists across repeated revisits. Existing systems either

safetyarxiv-cs-cv

New benchmark just dropped!

The pelican on a bicycle is sooo outdated, so I came up with a new, improved version. Qwen3.8-27b medium (UD-Q4_K_XL) vs. Sol 5.6 high vs. Qwen3.6-35B (UD-Q6_K_XL) Prompt (only real with typo!): 'Crea

Tatarstan Toponyms: A Bilingual Dataset and Hybrid RAG System for Geospatial Question Answering

arXiv:2605.05962v2 Announce Type: replace Abstract: This paper addresses end-to-end geospatial question answering over multilingual toponymic data. We introduce a bilingual (Russian-Tatar) dataset of

Training Chemical Plausibility-Aware Large Language Models for Single-Step Retrosynthesis

arXiv:2608.18940v1 Announce Type: cross Abstract: Single-step retrosynthesis is a central component of computer-aided synthesis planning, yet its intrinsically one-to-many nature is poorly captured by

A decodability criterion predicts when hidden-state selection beats majority voting in large language models

arXiv:2608.17124v1 Announce Type: new Abstract: Combining the answers a large language model (LLM) samples for a question into one decision is a test-time information fusion problem, usually solved by