AI 速递 2026-07-16
生成时间：2026/7/16 09:23:45（UTC: 2026-07-16T01:23:45.085Z）
Weak-to-Strong Generalization via Direct On-Policy Distillation
Section titled “Weak-to-Strong Generalization via Direct On-Policy Distillation”👍 112 · arXiv
Reinforcement learning with verifiable rewards (RLVR) is a powerful recipe for improving language-model reasoning, but it is expensive to repeat on every new strong model because the target model must…
ABot-N1: Toward a General Visual Language Navigation Foundation Model
Section titled “ABot-N1: Toward a General Visual Language Navigation Foundation Model”👍 84 · arXiv
Visual Language Navigation foundation models aim to unify deep reasoning for grounded spatial decisions with broad versatility for diverse embodied tasks. Current approaches typically achieve this int…
ABot-AgentOS: A General Robotic Agent OS with Lifelong Multi-modal Memory
Section titled “ABot-AgentOS: A General Robotic Agent OS with Lifelong Multi-modal Memory”👍 72 · arXiv
Recent VLM and VLA systems have improved robotic perception and action prediction, yet long-horizon embodied agents still require a general runtime layer for reasoning, memory, tool use, verification,…
Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks with Dense Reward-Based Grading
Section titled “Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks with Dense Reward-Based Grading”👍 64 · arXiv
AI agents have become capable of autonomously completing short, well-specified tasks. However, existing terminal benchmarks largely focus on simple problems that finish within minutes and are evaluate…
Video Generation Models are General-Purpose Vision Learners
Section titled “Video Generation Models are General-Purpose Vision Learners”👍 59 · arXiv
Driven by next-token prediction, NLP shifted from task-specific models into powerful generalist foundation models. What, then, is the equivalent catalyst needed to achieve a general-purpose model in c…
OpenClaw v2026.7.2-beta.1
Section titled “OpenClaw v2026.7.2-beta.1”2026.7.2
Section titled “2026.7.2”Highlights
Section titled “Highlights”- Remote coding sessions: run Control UI sessions on cloud workers, open Codex and Claude catalog sessions in terminals on their owning hosts, and resume OpenCode and …
链接：https://github.com/openclaw/openclaw/releases/tag/v2026.7.2-beta.1
vLLM v0.25.1
Section titled “vLLM v0.25.1”vLLM v0.25.1
Section titled “vLLM v0.25.1”Highlights
Section titled “Highlights”This release features 2 commits from 2 contributors (1 new)!
v0.25.1 is a patch release containing two targeted bug fixes on top of v0.25.0.
Bug Fixes
Section titled “Bug Fixes”- *…
链接：https://github.com/vllm-project/vllm/releases/tag/v0.25.1
Goose v1.43.0
Section titled “Goose v1.43.0”✨ Features
Section titled “✨ Features”- Reconnect desktop ACP sessions after sleep and connection loss #10411
- Per-message usage stats UI (tokens, cost, TTFT, tok/s) [#1021…
链接：https://github.com/aaif-goose/goose/releases/tag/v1.43.0
OpenAI Codex CLI rust-v0.145.0-alpha.15
Section titled “OpenAI Codex CLI rust-v0.145.0-alpha.15”Release 0.145.0-alpha.15
…
链接：https://github.com/openai/codex/releases/tag/rust-v0.145.0-alpha.15
Microsoft is reportedly training salespeople to talk down OpenAI and Anthropic
Section titled “Microsoft is reportedly training salespeople to talk down OpenAI and Anthropic”Microsoft is looking to sell its in-house AI models as more efficient and cost-effective than its competitors’ models.
Amid hardware legal battle, OpenAI releases a $230 keyboard for Codex
Section titled “Amid hardware legal battle, OpenAI releases a $230 keyboard for Codex”OpenAI, which is in the middle of a legal battle with Apple over hardware trade theft allegations, just released a light-up keyboard designed to be paired with its agentic coding app.
SpaceX falls to $135 IPO price ahead of Starship launch
Section titled “SpaceX falls to $135 IPO price ahead of Starship launch”The stock has steadily fallen from the euphoric post-IPO high, showing that markets may be sobering up to the promises CEO Elon Musk made before and after SpaceX went public.
Thinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling
Section titled “Thinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling”It’s the company’s first public proof point after a year and a half spent building AI infrastructure largely out of public view.
Hack suggests AI music generator Suno scraped YouTube for training data
Section titled “Hack suggests AI music generator Suno scraped YouTube for training data”The hacker used an employee’s credentials to access source code, which revealed how Suno scraped decades of audio.
Whatnot acquires Shaped to power real-time live shopping recommendations
Section titled “Whatnot acquires Shaped to power real-time live shopping recommendations”Livestream shopping platform Whatnot has acquired AI startup Shaped, a machine learning company focused on real-time recommendations and search. The deal will bolster Whatnot’s personalization and discovery features as it expands into new product categories.
Microsoft patches record number of security vulnerabilities, citing its use of AI
Section titled “Microsoft patches record number of security vulnerabilities, citing its use of AI”Microsoft’s monthly release of security fixes, dubbed Patch Tuesday, resolved a record 570 security vulnerabilities across the company’s product line, thanks to discoveries with AI.
Apple Intelligence approved for launch in China with Alibaba’s Qwen AI
Section titled “Apple Intelligence approved for launch in China with Alibaba’s Qwen AI”The deal, which was rumored to be in the works last year, marks an important step for Apple’s AI ambitions in a key market.