Crypto Briefing

MIT’s MeMo framework boosts LLM performance by 26% without retraining

MIT's MeMo framework trains a compact memory model that boosts LLM performance by up to 26.73% without retraining, with major implications for crypto AI agents.

Researchers automated LLM reasoning strategy design and cut token usage by 69.5%

Researchers from Meta and Google built AutoTTS to automatically discover optimal LLM reasoning strategies, cutting token ...
InfoWorld

Taming the generative AI back end

Writing code that interacts with LLM services requires bridging two different worlds. Use these tips and techniques to bind ...
IEEE

Building Energy Management Systems with LLM Agent Enhanced Natural Language Policy Explanation

Abstract: Building Energy Management System (BEMS) is important for optimizing energy usage. However, existing BEMS often lack natural language explanations, making it difficult for residents to ...
IEEE

Two-Step RAG for Metadata Filtering and Statistical LLM Evaluation

Abstract: This study addresses a limitation in Retrieval-Augmented Generation (RAG) systems: poor retrieval accuracy when vague prompts or metadata are missing. We propose the Two-Step RAG method to ...
GitHub

ToolSafe: Enhancing Tool Invocation Safety of LLM-based agents via Proactive Step-level Guardrail and Feedback

. ├── TS-Bench/ # Benchmark datasets for guardrail model evaluation ├── benchmark/ # Evaluation benchmark of agent safety&security ├── scripts/ # Shell scripts for training/inference ├── src/ # Source ...
Microsoft

LLM Reasoning as Trajectories: Step-Specific Representation Geometry and Correctness Signals

This work characterizes large language models’chain-of-thought generation as a structured trajectory through representation space. We show that mathematical reasoning traverses functionally ordered, ...