CAVE-Bench

What is CAVE-Bench?

LLM agents increasingly keep working after a task succeeds as they resume after compaction or take over handoffs.

Released

2026-09-26

Evaluates

General AI, Language & Knowledge, cs.AI

Openness

unknown

Importer

Claire Radar

Review status

rule-auto-admitted

Reported scores

0

Source provenance

• Original evidence https://arxiv.org/abs/2609.32616

CAVE-Bench paper

• Paper https://arxiv.org/abs/2609.32616
• Website https://henrymao2004.github.io/agent-over-correction/

Which sources cite CAVE-Bench?

• CAVE-Bench

Related General AI, Language & Knowledge, cs.AI benchmarks

• $\tau^\tau$-Bench: An Environment for End-To-End, Realistic Agent Construction
• 2d-fet-bench
• 3DCodeBench
• 5-Dialects-BN
• A Large-Scale Dataset and Benchmark
• A Real-World CAD Segmentation Benchmark
• A3S-Bench
• AARRI-Bench