Research map / Large language models / Language, safety and interpretability
Not All Answers Are Contextually Persuadable: Inference Dynamics in Large Language Models under Contextual Influence
Zongye Hu, Weiqing Luo, Yanjie Fu, Yu Gan, Haofeng Zhang, Ziyi Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| tighten convergence tolerance from 0.1 to 0.01 | Qwen2.5-1.5B attention layers · OpenBookQA, N=1000 | % converged layers: 93.0 -> 47.6 | Hurt |
| tighten convergence tolerance from 0.1 to 0.01 | Falcon3-3B-Base attention layers · MINTAKA, N=1000 | % converged layers: 100 -> 93.3 | Hurt |
| tighten convergence tolerance from 0.1 to 0.01 | Qwen3-4B attention layers · Simple QA, N=1000 | % converged layers: 100 -> 92.7 | Hurt |
| increase repetition length N from 500 to 1000 | answer-shift sign prediction, Transfer regime · Transfer (mean of 6 models) | accuracy: 85.5 -> 86.1 | No clear effect |
| increase repetition length N from 500 to 1000 | answer-shift sign prediction, Mislead regime · Mislead (mean of 6 models) | accuracy: 82.6 -> 80.0 | Hurt |
| increase repetition length N from 500 to 1000 | answer-shift sign prediction, Correct regime · Correct (mean of 6 models) | F1: 78.9 -> 89.8 | Improved |
| compare predicted shift magnitude with Amp across models | Qwen2.5-1.5B · SYCON | stability agreement: 4.25 -> 0.13 | Mixed |
Closest papers by meaning
- A Simple Model of Inference Scaling Laws (ICML 2025 · similarity 0.88)
- In-Context Learning Dynamics with Random Binary Sequences (ICLR 2024 · similarity 0.87)
- Unveiling and Manipulating Prompt Influence in Large Language Models (ICLR 2024 · similarity 0.86)
- Supervised Knowledge Makes Large Language Models Better In-context Learners (ICLR 2024 · similarity 0.86)
- Taming Knowledge Conflicts in Language Models (ICML 2025 · similarity 0.86)
- Evolving Prompts In-Context: An Open-ended, Self-replicating Perspective (ICML 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.
