← all datasets

JSON

Emerging
3papers using it
2025first seen

The 'JSON' dataset/benchmark contains structured data formatted in JavaScript Object Notation and is used to evaluate the effectiveness of lossy text compression methods by measuring how well an independent decoder LLM can reconstruct semantic information from compressed representations.

Papers using JSON (3)

JSON β€” datasets β€” llm-papers