Every story we have kept, newest first.
Looking for the daily editions? → Past editions
Page 1331 of 1373
AI agents excel on benchmarks but fail in practice due to single-metric evaluations ignoring consistency, robustness, predictability, and safety. This arXiv paper proposes 12 concrete metrics across these four dimensions, grounded in safety-critical engineering.
ArXiv AI · 211d ago
Proxy State-Based Evaluation introduces an LLM-driven simulation framework for benchmarking multi-turn tool-calling agents, avoiding costly deterministic backends. It uses scenarios to define goals and states, with LLM trackers inferring proxy states from traces for verification.
ArXiv AI · 211d ago
PAHF is a framework for continual personalization of AI agents, learning online from live human interactions via explicit per-user memory. It uses a three-step loop: pre-action clarification, preference-grounded actions, and post-action feedback for memory updates.
ArXiv AI · 211d ago
January Mirror, an evidence-grounded clinical reasoning system, scored 87.5% on a 120-question 2025 endocrinology board-style exam, outperforming human experts (62.3%) and frontier LLMs like GPT-5.2 (74.6%). It excelled on the hardest questions (76.7% accuracy) under closed-evidence constraints without web access.
ArXiv AI · 211d ago
Researchers demonstrate that sequence models' in-context learning induces cooperation in multi-agent RL without hardcoded co-player assumptions or timescale separation. Training against diverse co-players leads to best-response strategies on intra-episode timescales.
ArXiv AI · 211d ago
Researchers launch GPSBench, a 57,800-sample dataset across 17 tasks to probe LLMs' geospatial reasoning without tools. 14 SOTA LLMs show reliability in geographic knowledge but struggle with geometric computations like distance and bearing.
ArXiv AI · 211d ago
FoT introduces a general-purpose framework for dynamic reasoning schemes in LLMs, overcoming static structures in Chain of Thought, Tree of Thoughts, and Graph of Thoughts. It features hyperparameter tuning, prompt optimization, parallel execution, and caching for better performance.
ArXiv AI · 211d ago
Surge AI launches Corecraft, the first high-fidelity RL environment in EnterpriseGym, simulating enterprise customer support with 2,500+ entities and 23 tools. Training GLM 4.6 via GRPO improves task pass@1 from 25% to 37% on held-out tasks, with gains transferring to BFCL (+4.5%), τ²-Bench Retail (+7.4%), and Toolathlon (+6.8%).
ArXiv AI · 211d ago
Neural solvers excel in simple routing but falter on complex constraints. CaR introduces the first general framework using explicit learning-based feasibility refinement and joint training to generate diverse solutions for lightweight improvement.
ArXiv AI · 211d ago
CAFE reformulates automated feature engineering as a causally-guided sequential decision process using causal discovery for soft priors and multi-agent RL for construction. It outperforms baselines by up to 7% on 15 benchmarks and reduces performance drops 4x under covariate shifts.
ArXiv AI · 211d ago
Proposes a trainable framework for interactive in-context learning using multi-turn feedback from information asymmetry on verifiable tasks. Trained smaller models nearly match performance of 10x larger models and generalize to coding, puzzles, and mazes.
ArXiv AI · 211d ago
This arXiv paper proposes a 'store then on-demand extract' approach for AI memory to retain raw experiences and avoid information loss from the dominant 'extract then store' method. It also explores deriving deeper insights from probabilistic experiences and improving efficiency via shared storage.
ArXiv AI · 211d ago
Autonomous agentic workflows exhibit optimization instability, where iterative self-improvement degrades classifier performance, especially for low-prevalence clinical symptoms like Long COVID brain fog (3%). Using the open-source Pythia framework, validation sensitivity oscillated wildly between 1.0 and 0.0.
ArXiv AI · 211d ago
ArXiv paper defines Agent Skill framework mathematically and evaluates its benefits for small language models (SLMs) in industrial settings with data security constraints. Moderate SLMs (12B-30B params) show substantial gains in accuracy and reduced hallucinations; 80B code-specialized variants match proprietary models with better GPU efficiency.
ArXiv AI · 211d ago
Morgan Stanley identifies 2026 as a pivotal inflection point for China's humanoid robot market, mirroring the 2019-2020 NEV surge. IDC predicts application scenarios will triple by 2026.
cnBeta (Full RSS) · 211d ago
Tesla announced FSD Supervised cumulative mileage exceeds 8 billion miles, up from 7 billion in December 2024. This data accelerates training for unsupervised Full Self-Driving.
IT之家 · 211d ago
Ant Group launches a trillion-parameter open-source model superior in human understanding and execution. It excels in emotional intelligence and agent combat power.
量子位 · 211d ago
After China's Spring Festival Gala (Chunwan), various AI systems and robots have converged to a single location. This place has captured the 'stray' elements from the gala, hinted at with a doge meme.
量子位 · 211d ago
After 2026 Chunwan, AI products like Tencent Yuanbao and robots from Songyan Dynamics and Unitree shifted to Bilibili for live streams and interactions to sustain hype. Bilibili's interest rooms and AI community enabled peaks of 1M+ viewers.
虎嗅 · 211d ago
Chinese giants invest over 45B RMB in AI voice payment subsidies during 2026 Spring Festival to seize user intents via agents. ByteDance's Doubao AI phone was blocked by rivals' ecosystems, while Alibaba's Qwen thrives in closed-loop integration across its services.
虎嗅 · 211d ago
Telstra has identified 380 potential AI use cases across its operations. The company warns that AI cost-benefits require close examination to ensure viable implementations.
iTNews Australia · 211d ago
HBR study reveals AI intensifies work by blurring boundaries, expanding tasks, and eliminating rest gaps. Employees face workload creep, role blurring, and 'speed tyranny' without time savings.
虎嗅 · 211d ago
OpenAI is partnering with India's Tata Group on AI technologies. The collaboration includes building data center infrastructure.
Bloomberg Technology · 211d ago
Indian Prime Minister Narendra Modi and French President Emmanuel Macron will speak at India's AI Summit. Global tech leaders are also slated to attend the event.
Bloomberg Technology · 211d ago
Chinese AI and robotics firms like Tencent and AgiBot are appointing millennials and Gen Z talents as chief scientists to spearhead AI and robotics research. Highlighted is 28-year-old Vinces Yao Shunyu, former OpenAI researcher, who joined Tencent as chief AI scientist reporting directly to President Martin Lau.
SCMP Technology · 211d ago
Amazon's generative AI-powered next-gen assistant Alexa+ is rolling out in the US. Amazon Japan revealed at an Echo Dot event that Japan is the next priority market after US and Canada, but no specific timeline was provided.
ITmedia AI+ (日本) · 211d ago
Local AI agent OpenClaw drives Raspberry Pi stock nearly double in 3 days to £1B cap on cheap hardware hype. Runs on $100 Pi for secure edge tasks vs cloud or pricey Macs.
虎嗅 · 211d ago
Taichu Yuangi completed deep adaptations of key Chinese open-source LLMs, including Zhipu GLM-5.0 and Alibaba Qwen3.5-397B-A17B, on its self-developed T100 accelerator card. The SDAA software stack introduces a tiered developer toolchain addressing varied skill levels.
36氪 · 211d ago
Microsoft's Project Silica etches data into glass with femtosecond lasers, packing 4.84TB into a 12cm square stable over 10,000 years at room temperature. Reading uses microscopes and CNNs; writing is slow at 8MB/s, taking 150+ hours per plate.
IT之家 · 211d ago