Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
Large Language Models Cannot Self-Correct Reasoning Yet
2024-01-16
· ICLR 2024 poster ·
anchor
Findings
IC-852
Intrinsic self-correction without external feedback consistently degrades reasoning accuracy across GPT-3.5-turbo, GPT-4, GPT-4-turbo, and LLaMA-2-70B-chat
IC-853
Multi-agent debate with GPT-3.5-turbo-0301 does not outperform self-consistency at equivalent inference cost on GSM8K
IC-854
The apparent self-correction improvement in constrained generation (Madaan et al., 2023) is an artefact of a sub-optimal initial prompt, not a genuine model capability