Posts

All the articles I've posted.

5.0
Hands on with X's new AI-powered custom feeds
2026年04月23日
· TechCrunch· 04/23 08:00 采集
X 推出 AI 驱动的自定义信息流替代 Communities，由 Grok 策划内容并引入新广告位。
7.0
Sam Altman's Orb Company Promoted a Bruno Mars Partnership That Doesn't Exist
2026年04月23日
· Wired· 04/23 08:00 采集
Sam Altman 的 Orb 公司宣传了与 Bruno Mars 的合作，但艺人方表示从未有过任何相关讨论。
7.5
5 AI Models Tried to Scam Me. Some of Them Were Scary Good
2026年04月23日
· Wired· 04/23 08:00 采集
实测 5 个 AI 模型的钓鱼和社会工程攻击能力，部分模型的表现令人担忧。
7.0
AI Tools Are Helping Mediocre North Korean Hackers Steal Millions
2026年04月23日
· Wired· 04/23 08:00 采集
朝鲜黑客组织利用 AI 工具编写恶意软件、创建虚假公司网站，三个月内窃取高达 1200 万美元。
6.5
AI failure could trigger the next financial crisis, warns Elizabeth Warren
2026年04月23日
· arXiv· 04/23 08:00 采集
7.0
OpenAI now lets teams make custom bots that can do work on their own
2026年04月23日
· arXiv· 04/23 08:00 采集
7.5
Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts
2026年04月23日
· arXiv· 04/23 08:00 采集
6.5
Accelerating PayPal's Commerce Agent with Speculative Decoding: An Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models
2026年04月23日
· arXiv· 04/23 08:00 采集
6.5
Statistics, Not Scale: Modular Medical Dialogue with Bayesian Belief Engine
2026年04月23日
· arXiv· 04/23 08:00 采集
6.5
Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders
2026年04月23日
· arXiv· 04/23 08:00 采集
7.0
Continuous Semantic Caching for Low-Cost LLM Serving
2026年04月23日
· arXiv· 04/23 08:00 采集
6.0
On the Quantization Robustness of Diffusion Language Models in Coding Benchmarks
2026年04月23日
· arXiv· 04/23 08:00 采集
7.5
TTKV: Temporal-Tiered KV Cache for Long-Context LLM Inference
2026年04月23日
· arXiv· 04/23 08:00 采集
7.0
Temporally Extended Mixture-of-Experts Models
2026年04月23日
· arXiv· 04/23 08:00 采集
6.0
Scaling Self-Play with Self-Guidance
2026年04月23日
· arXiv· 04/23 08:00 采集
6.0
R2IF: Aligning Reasoning with Decisions via Composite Rewards for Interpretable LLM Function Calling
2026年04月23日
· arXiv· 04/23 08:00 采集
6.5
Efficient Test-Time Inference via Deterministic Exploration of Truncated Decoding Trees
2026年04月23日
· arXiv· 04/23 08:00 采集
6.5
GRPO-VPS: Enhancing Group Relative Policy Optimization with Verifiable Process Supervision for Effective Reasoning
2026年04月23日
· arXiv· 04/23 08:00 采集
6.5
MGDA-Decoupled: Geometry-Aware Multi-Objective Optimisation for DPO-based LLM Alignment
2026年04月23日
· arXiv· 04/23 08:00 采集
7.0
Supplement Generation Training for Enhancing Agentic Task Performance
2026年04月23日
· arXiv· 04/23 08:00 采集

Posts

Hands on with X's new AI-powered custom feeds

Sam Altman's Orb Company Promoted a Bruno Mars Partnership That Doesn't Exist

5 AI Models Tried to Scam Me. Some of Them Were Scary Good

AI Tools Are Helping Mediocre North Korean Hackers Steal Millions

AI failure could trigger the next financial crisis, warns Elizabeth Warren

OpenAI now lets teams make custom bots that can do work on their own

Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts

Accelerating PayPal's Commerce Agent with Speculative Decoding: An Empirical Study on EAGLE3 with Fine-Tuned Nemotron Models

Statistics, Not Scale: Modular Medical Dialogue with Bayesian Belief Engine

Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders

Continuous Semantic Caching for Low-Cost LLM Serving

On the Quantization Robustness of Diffusion Language Models in Coding Benchmarks

TTKV: Temporal-Tiered KV Cache for Long-Context LLM Inference

Temporally Extended Mixture-of-Experts Models

Scaling Self-Play with Self-Guidance

R2IF: Aligning Reasoning with Decisions via Composite Rewards for Interpretable LLM Function Calling

Efficient Test-Time Inference via Deterministic Exploration of Truncated Decoding Trees

GRPO-VPS: Enhancing Group Relative Policy Optimization with Verifiable Process Supervision for Effective Reasoning

MGDA-Decoupled: Geometry-Aware Multi-Objective Optimisation for DPO-based LLM Alignment

Supplement Generation Training for Enhancing Agentic Task Performance