Research map / Large language models / Language, safety and interpretability
Natural Language Inference in Context – Investigating Contextual Reasoning over Long Texts
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Fine-tune BART-NLI on ConTRoL after training on SNLI, MultiNLI, and Adversarial NLI | BART-NLI · ConTRoL test | accuracy: 45.02 -> 60.95 | Helped |
| Increase context length | BART on ConTRoL · ConTRoL, contexts shorter than 500 words versus longer than 3,000 words | accuracy: 65% -> 40% | Hurt |
| Use the hypothesis without the context (ablation) | BERT on ConTRoL · ConTRoL test | accuracy: 50.62 -> 36.07% | Hurt |
| Use the hypothesis without the context (ablation) | BART on ConTRoL · ConTRoL test | accuracy: 56.34 -> 43.41% | Hurt |
| Use the context without the hypothesis (ablation) | BERT on ConTRoL · ConTRoL test | accuracy: 50.62 -> 33.09% | Hurt |
| Use the context without the hypothesis (ablation) | BART on ConTRoL · ConTRoL test | accuracy: 56.34 -> 35.94% | Hurt |
| Evaluate on multi-paragraph rather than single-paragraph instances | BERT on ConTRoL · ConTRoL, multi-paragraph versus single-paragraph | accuracy: 51.17% -> 40.30% | Hurt |
Closest papers by meaning
- A Controllable Examination for Long-Context Language Models (NeurIPS 2025 · similarity 0.88)
- Hypothesis Search: Inductive Reasoning with Language Models (ICLR 2024 · similarity 0.86)
- NoLiMa: Long-Context Evaluation Beyond Literal Matching (ICML 2025 · similarity 0.86)
- Longer Context, Deeper Thinking: Uncovering the Role of Long-Context Ability in Reasoning (NeurIPS 2025 · similarity 0.86)
- BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystack (NeurIPS 2024 · similarity 0.86)
- Linguini: A benchmark for language-agnostic linguistic reasoning (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.