Nov 2, 2025

The Smol Training Playbook

The Smol Training Playbook reveals essential strategies for developing world-class LLMs, emphasizing practical techniques and community-driven insights from the Hugging Face ecosystem.

Tongyi DeepResearch – open-source 30B MoE Model that rivals OpenAI DeepResearch

Tongyi DeepResearch emerges as the first fully open-source Web Agent, achieving state-of-the-art performance with scores of 32.9 on Humanity’s Last Exam and 75 on xbench-DeepSearch, surpassing all existing agents.

Policy, privacy and post-quantum: anonymous credentials for everyone

Anonymous credentials (AC) offer a method to rate-limit AI agents without compromising user privacy, allowing for fine-grained security policies that adapt to the evolving landscape of web traffic driven by AI.

Context engineering

Context engineering evolves from traditional prompt engineering, emphasizing a dynamic and structured approach to inputting tokens into LLMs, enhancing their decision-making capabilities in complex systems.

Is Your Bluetooth Chip Leaking Secrets via RF Signals?

Bluetooth chips are vulnerable to machine learning-assisted side-channel attacks, allowing attackers to recover full encryption keys from 90,000 traces captured at a distance of one meter, affecting millions of devices globally, including wearables and IoT products.

[R] TempoPFN: Synthetic Pretraining of Linear RNNs for Zero-Shot Timeseries Forecasting

TempoPFN is a univariate time series foundation model that excels in zero-shot forecasting by leveraging synthetic data, utilizing a GatedDeltaProduct architecture for enhanced performance and efficiency.

[P] Beyond Simple Retrieval — Smarter Context for Smarter LLMs

[P] Introducing Hephaestus: AI workflows that build themselves as agents discover what needs to be done