AI Glossary

What is Tool Use?

Overview

Tool use is the ability of an AI system to call external tools such as search, code execution, databases, calculators, or business APIs. It matters because many real tasks require current data or side effects that a language model cannot provide from weights alone.

Why it matters

Tool use is a core capability for agents that need to inspect, decide, and act across software environments.

Where it appears in AI research

Agent benchmarks
Coding assistant workflows
Enterprise AI automation
Function calling and MCP integrations

Related terms

Function Calling MCP Agent Evaluation

Related DeepSignal articles

arXiv cs.CL·Leyao Wang, Yanan He, Peng Chen, Asaf Yehudai, Yixin Liu, Rex Ying, Michal Shmueli-Scheuer, Arman Cohan

2w ago

FeaturedOriginal

Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?

AI Summary

The REFLECT benchmark reveals that current LLM judges are unreliable, achieving below 55% accuracy in evaluating reasoning and evidence use, highlighting the need for improved evaluation methods for deep research agents.

#LLM #Agent #Inference #Policy

2

arXiv cs.CL·Haiquan Lu, Zigeng Chen, Gongfan Fang, Xinyin Ma, Xinchao Wang

2w ago

FeaturedOriginal

Mix-Quant: Quantized Prefilling, Precise Decoding for Agentic LLMs

AI Summary

Mix-Quant introduces a phase-aware quantization framework for agentic LLMs, achieving up to 3x speedup in the prefilling stage while maintaining BF16 precision during decoding. This approach alleviates the computational bottleneck in long-context, multi-turn inference, demonstrating significant efficiency improvements without substantial accuracy loss.

#LLM #Agent #Inference

4

arXiv cs.CL·Mehrdad Saberi, Keivan Rezaei, Soheil Feizi

2w ago

FeaturedOriginal

SpecHop: Continuous Speculation for Accelerating Multi-Hop Retrieval Agents

AI Summary

SpecHop introduces a continuous speculation framework for multi-hop retrieval tasks, reducing latency by up to 40% while maintaining accuracy. By leveraging multiple speculative threads and asynchronous verification, it approaches oracle latency gains, significantly enhancing the efficiency of large language models in information-intensive applications.

#Agent #Inference #AI Startup

1

arXiv cs.CL·Juan S. Santillana

3w ago

FeaturedOriginal

VectraYX-Nano: A 42M-Parameter Spanish Cybersecurity Language Model with Curriculum Learning and Native

AI Summary

VectraYX-Nano is a 41.95M-parameter Spanish cybersecurity language model that utilizes curriculum learning and native tool invocation, achieving a conversational gate of 0.78 after extensive training on a diverse 170M-token corpus. This model is the first of its kind with end-to-end integration and operates efficiently on commodity hardware.

#LLM #Security #AI Startup

6

Anthropic scales Project Glasswing to 150 partners across 15 countries to hunt critical software flaws

The Decoder·Matthias Bastian

4d ago

FeaturedOriginal

Anthropic scales Project Glasswing to 150 partners across 15 countries to hunt critical software flaws

AI Summary

Anthropic has expanded Project Glasswing to 150 partners in over 15 countries, utilizing Claude Mythos Preview to identify critical software vulnerabilities. This initiative has already uncovered over 10,000 serious flaws, while Anthropic simultaneously markets Claude Security for commercial fixes, effectively profiting from both vulnerability detection and remediation.

#Security #AI Startup #Policy

1

Overview

Why it matters

Where it appears in AI research

Related terms

Related DeepSignal articles

Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?

Mix-Quant: Quantized Prefilling, Precise Decoding for Agentic LLMs

SpecHop: Continuous Speculation for Accelerating Multi-Hop Retrieval Agents

VectraYX-Nano: A 42M-Parameter Spanish Cybersecurity Language Model with Curriculum Learning and Native Tool Use

Anthropic scales Project Glasswing to 150 partners across 15 countries to hunt critical software flaws

VectraYX-Nano: A 42M-Parameter Spanish Cybersecurity Language Model with Curriculum Learning and Native