Introduces IFBench benchmark with 58 new constraints and demonstrates RLVR training improves generalization of language models to unseen verifiable output constraints.
Gsm-symbolic: Understanding the limitations of mathematical rea- soning in large language models
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
-
Generalizing Verifiable Instruction Following
Introduces IFBench benchmark with 58 new constraints and demonstrates RLVR training improves generalization of language models to unseen verifiable output constraints.