Research map / Large language models / Alignment and preferences
CoP: Factual Inconsistency Detection by Controlling the Preference
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 |
|---|---|---|---|
| Use prompted extra inference to compare generation probabilities | CoP factual inconsistency detection · XSum Hallucination Annotations | F1: 58.90 -> 63.54 | Helped |
| Tune the prompt using few-shot labeled data | CoP · XSum Hallucination Annotations corpus | Corpus F1: 63.72 -> 66.56 | Helped |
| Tune the prompt using full-shot labeled data | CoP · XSum Hallucination Annotations corpus | Corpus F1: 63.72 -> 69.61 | Helped |
| Add entity tokens to the prompt (ablation) | CoP · FRANKCNN EntE | Pearson correlation: 33.32 -> 34.66 | Helped |
| Add entity tokens to the prompt (ablation) | CoP · FRANKCNN OutE | Pearson correlation: 23.47 -> 26.09 | Helped |
| Add resolved reference tokens to the prompt (ablation) | CoP · FRANKCNN CorefE | Pearson correlation: 34.77 -> 35.47 | Helped |
| Add resolved reference tokens to the prompt (ablation) | CoP · FRANKCNN OutE | Pearson correlation: 37.89 -> 36.55 | Hurt |
| Increase prompt-vector length (ablation) | CoP prompt tuning · XSum Hallucination Annotations | Corpus F1 | Mixed |
Closest papers by meaning
- Do Automatic Factuality Metrics Measure Factuality? A Critical Evaluation (NeurIPS 2025 · similarity 0.85)
- PRISMM-Bench: A Benchmark of Peer-Review Grounded Multimodal Inconsistencies (ICLR 2026 · similarity 0.83)
- To Believe or Not to Believe Your LLM: Iterative Prompting for Estimating Epistemic Uncertainty (NeurIPS 2024 · similarity 0.83)
- Alignment Imprint: Zero-Shot AI-Generated Text Detection via Provable Preference Discrepancy (NeurIPS 2026 · similarity 0.83)
- Precise Information Control in Long-Form Text Generation (NeurIPS 2025 · similarity 0.82)
- Fine-Tuning Language Models for Factuality (ICLR 2024 · similarity 0.82)
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.