AI news today

This source gives a title and a link, nothing to quote.

They claim — We evaluated 11 widely used open-source ASR models and found that several of the highest-scoring systems reproduced benchmark transcripts from the VoxPopuli English and LibriSpeech (clean…

Published 2026-08-21

Fetched 2026-08-25 10:14 UTC · 8de96fd631a45d0c

Torn from — the page itself

“InfinityEdit enables continuous, unbounded video editing by using a lightweight adapter with history, temporal causal, and edit…”

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · 006f6bff02cfdc40

Torn from — the abstract

“Graph Engineering organizes multi-agent LLM systems through dynamic graph structures to coordinate specialized agents and manag…”

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · 34bb2eff51f8481e

Torn from — the abstract

“OmniAssistBench evaluates real-time interactive video assistants by reverse-engineering multi-turn interaction videos, revealin…”

They claim — Results show that the proprietary Gemini-3-Pro reaches 66.4 out of the max point of 100, while the open-source Qwen3-Omni-Instruct achieves 51.2.

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · cb7314dbba846623

Torn from — the abstract

“TLive-Omni is an omni-modal model for live-commerce that unifies image, video, audio, and text via timestamped token grouping, …”

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · 78c306c48010b108

Torn from — the abstract

This source gives a title and a link, nothing to quote.

They claim — Faster inference : up to 3.18 throughput improvement on a GPU and up to 2.87x on-device.

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · b41fa9c74f86007c

Torn from — the page itself

“Provider-neutral Codex Skill for producing verified AI presenter videos from a script and an authorized presenter image.”

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · f2ced8b6406f4329

What the account wrote — The account left it blank

Torn from — the README

“A two-step hyperparameter transfer framework predicts optimal learning rates for large Mixture-of-Experts models by scaling acr…”

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · 7e8aaa6cb8c41512

Torn from — the abstract

“Introducing AI Futures, a new OpenAI blog exploring how transformative AI could reshape power, governance, the economy, and ind…”

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · ec2332dc6523ff28

Torn from — the listing only (the original would not open)

“With a fixed deadline and design resources committed elsewhere, Stampli used Codex and ChatGPT Work to compress weeks of launch…”

Published 2026-08-20

Fetched 2026-08-25 10:14 UTC · 1476f459afd654a9

Torn from — the listing only (the original would not open)

“SWE-bench Science benchmarks coding agents on scientific software repair, revealing failure mechanisms and mixed effects of sci…”

They claim — Even the best-performing agent, Claude Code with Opus-5 (max), achieves a pass@1 below 50%, highlighting the substantial challenges posed by scientific software engineering.

Published 2026-08-19

Fetched 2026-08-25 10:14 UTC · bbf6eac293bcc02d

Torn from — the abstract

This source gives a title and a link, nothing to quote.

They claim — A few subjects like philosophy and European history actually improved (+6pp and +4pp respectively), while abstract algebra and formal logic saw larger drops.

Published 2026-08-19

Fetched 2026-08-25 10:14 UTC · 30b05f2788443301

Torn from — the model card

This source gives a title and a link, nothing to quote.

They claim — More efficient native 4K generation: more coherent global structure, color harmony, and stable high-resolution output with improved generation efficiency.

Published 2026-08-19

Fetched 2026-08-25 10:14 UTC · dc5f49c564074f43

Torn from — the model card

This source gives a title and a link, nothing to quote.

Published 2026-08-19

Fetched 2026-08-25 10:14 UTC · e0cf23a392b802ca

Torn from — the model card

This source gives a title and a link, nothing to quote.

They claim — 🐦 We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement.

Published 2026-08-19

Fetched 2026-08-25 10:14 UTC · 4195fce25151562f

Torn from — the model card

This source gives a title and a link, nothing to quote.

They claim — 🐦 We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement.

Published 2026-08-18

Fetched 2026-08-25 10:14 UTC · 435a2ec6edeb1ae4

Torn from — the model card

This source gives a title and a link, nothing to quote.

Published 2026-08-18

Fetched 2026-08-25 10:14 UTC · c48c76b964ad9c2d

Torn from — the model card

“Sprix AI at 屿智同行 — state-aware SELF/COLLABORATE/HANDOFF routing for A2A agent networks.”

They claim — RouteLLM: Learning to Route LLMs with Preference Data, ICLR 2025 — preference-based cost-quality routing.

Published 2026-08-18

Fetched 2026-08-25 10:14 UTC · 1c90000bebc39c3d

What the account wrote — The account left it blank

Torn from — the README

This source gives a title and a link, nothing to quote.

Published 2026-08-18

Fetched 2026-08-25 10:14 UTC · 7e63adaac2fe5407

Torn from — the listing only (the original would not open)

“Autoprompt is a coding-agent skill that cuts failures by 45% on agentic coding tasks.”

They claim — Autoprompt is a coding-agent skill that cuts failures by 45% on agentic coding tasks.

Published 2026-08-17

Fetched 2026-08-25 10:14 UTC · 5d07e174c6aaeff9

What the account wrote — DE

Torn from — the README

“ParaTempo improves parallel reasoning efficiency by using temporal confidence to dynamically prune, retire, and reallocate reas…”

They claim — Experiments on challenging mathematical and scientific reasoning benchmarks show that ParaTempo reduces average latency by 21.8-32.2% and total token usage by 18.1-30.3% while maintaining…

Published 2026-08-17

Fetched 2026-08-25 10:14 UTC · 0f7a16969a7c76aa

Torn from — the abstract

“Hybrid-thinking multimodal language models suffer from response-pattern misalignment between thinking and non-thinking modes, w…”

Published 2026-08-16

Fetched 2026-08-25 10:14 UTC · 6a62c60a4eba3ed1

Torn from — the abstract

“A hierarchical taxonomy and dense supervision strategy improve diffusion-based image editing through fine-grained concepts, lar…”

They claim — To address these issues, we establish a comprehensive hierarchical taxonomy featuring over 1,000 fine-grained edit concepts and build ConceptEdit-12M, a massive dataset of 12 million high…

Published 2026-08-16

Fetched 2026-08-25 10:14 UTC · 38d02ea00a52b522

Torn from — the abstract

“dsh-routing-suite — injector + router-standard kit: install the runtime injector first, then the task-aware reasoning-mode ro…”

Published 2026-08-14

Fetched 2026-08-25 10:14 UTC · 5a272e5aab7c2bef

What the account wrote — 天津

Torn from — the README

“通过扫码或机器人凭据把IM机器人接入DeepSeek Harness（支持飞书、微信、钉钉、企业微信、QQ、Slack…”

Published 2026-08-14

Fetched 2026-08-25 10:14 UTC · 6e0d29f870534d9f

What the account wrote — The account left it blank

Torn from — the README

This source gives a title and a link, nothing to quote.

Published 2026-08-12

Fetched 2026-08-25 10:14 UTC · cccf20cd52859c98

Torn from — the model card

“AgentTeams plugin for DeepSeek Harness”

Published 2026-08-12

Fetched 2026-08-25 10:14 UTC · ed5787155a705d01

What the account wrote — ShenZhen

Torn from — the README

This source gives a title and a link, nothing to quote.

Published 2026-08-19

Fetched 2026-08-20 07:11 UTC · 0d3b50995f0f64d4

Torn from — the page itself

“Here’s how you can use Google Search tools to stu…”

Published 2026-08-19

Fetched 2026-08-20 07:11 UTC · 1b0691e082f0e950

Torn from — the page itself

“OpenAI reaffirms Zero Data Retention for eligible…”

Published 2026-08-19

Fetched 2026-08-20 07:11 UTC · 971da21a3ac32a67

Torn from — the listing only (the original would not open)