The Martian Apart

LLM Fingerprinting Through Semantic Variability & Cognitive Load Analysis

270
Total API Requests
96%
Technical Jargon
Drop Rate
0
Tool Hallucinations
Detected
5
Distraction
Hypotheses Tested

๐Ÿ”ง Tool Intent Analysis

Discover how semantic noise affects tool selection without causing hallucinations. Models gracefully degrade under cognitive load, dropping optional features while maintaining core functionality.

๐ŸŽฏ Overview: Models Don't Hallucinate

๐Ÿ” Discovery: But They Do Simplify

๐Ÿงช Experiment: What Causes Dropping?

๐Ÿ”ฌ Technical Deep Dive

๐Ÿ“ Data Access