Long-Horizon-Terminal-Bench 2.0: Automating Diverse Task Construction and Measuring Sustained Agent Progress

Not provided

benchmarkingtask constructionagent evaluation

2610.14146

Builder Relevance

80%

2h ago

Abstract

LHTB 2.0 extends the original benchmark to measure sustained agent progress through diverse tasks across multiple domains.

Reality Card

Core Claim

LHTB 2.0 provides a broader and more process-aware testbed for evaluating agent execution over extended periods, correcting previous defects and extending tasks.

Method / Result

LHTB 2.0 includes 77 tasks across 13 domains, with a significant increase in evaluation horizon from 90 minutes to three hours.

Limitations

The reliance on specific parameterized reference implementations may limit generalizability and reproducibility of results.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.

← Back to all papers