LRMs underperform on simple system 1 questions in both accuracy and efficiency, with problem difficulty implicitly encoded in early hidden states.
Reasoning processes that only indirectly support the Ground Truth or result in partially aligned conclusions should be excluded
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Exploring the System 1 Thinking Capability of Large Reasoning Models
LRMs underperform on simple system 1 questions in both accuracy and efficiency, with problem difficulty implicitly encoded in early hidden states.