FOL-Traces: Verified First-Order Logic Reasoning Traces at Scale
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866908784080715776 |
|---|---|
| author | Lee, Isabelle Liaw, Sarah Yogatama, Dani |
| author_facet | Lee, Isabelle Liaw, Sarah Yogatama, Dani |
| contents | Reasoning in language models is difficult to evaluate: natural-language traces are unverifiable, symbolic datasets are too small, and most benchmarks conflate heuristics with inference. We present FOL-Traces, the first large-scale dataset of programmatically verified reasoning traces, enabling rigorous evaluation of structured logical inference. We also propose two challenging and comprehensive diagnostic tasks-masked operation prediction and step completion-that directly probe syntactic awareness and process fidelity. FOL-Traces serves as a scalable testbed for rigorously studying how models perform structured logical inference. Systematic experiments with 5 reasoning LLMs show that the dataset remains challenging: models only reach around 45.7% accuracy on masked operation prediction and around 27% on two-step completion. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_14932 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | FOL-Traces: Verified First-Order Logic Reasoning Traces at Scale Lee, Isabelle Liaw, Sarah Yogatama, Dani Artificial Intelligence Reasoning in language models is difficult to evaluate: natural-language traces are unverifiable, symbolic datasets are too small, and most benchmarks conflate heuristics with inference. We present FOL-Traces, the first large-scale dataset of programmatically verified reasoning traces, enabling rigorous evaluation of structured logical inference. We also propose two challenging and comprehensive diagnostic tasks-masked operation prediction and step completion-that directly probe syntactic awareness and process fidelity. FOL-Traces serves as a scalable testbed for rigorously studying how models perform structured logical inference. Systematic experiments with 5 reasoning LLMs show that the dataset remains challenging: models only reach around 45.7% accuracy on masked operation prediction and around 27% on two-step completion. |
| title | FOL-Traces: Verified First-Order Logic Reasoning Traces at Scale |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2505.14932 |