Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
ANCHOR: Abductive Network Construction with Hierarchical Orchestration for Reliable Probability Inference in Large Language Models
Wentao Qiu, Guanran Luo, Zhongquan Jian, Jingqi Gao, Meihong Wang, Qingqiang Wu
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 |
|---|---|---|---|
| add causal Bayesian network to Naive Bayes in the aggregation (ablation) | Anchor (Qwen2.5-32B) · Common2Sense | Average F1: 0.558 -> 0.572 | Improved |
| add causal Bayesian network to Naive Bayes in the aggregation (ablation) | Anchor (Qwen2.5-72B) · Common2Sense | Average F1: 0.506 -> 0.596 | Improved |
| add Naive Bayes to the CBN in the aggregation (ablation) | Anchor (Qwen2.5-32B) · Common2Sense | Average F1: 0.479 -> 0.572 | Improved |
| add Naive Bayes to the CBN in the aggregation (ablation) | Anchor (Qwen2.5-72B) · Common2Sense | Average F1: 0.557 -> 0.596 | Improved |
| add hierarchical clustering of the expanded factor space (ablation) | Anchor (Qwen2.5-72B) · Common2Sense | Average F1: 0.400 -> 0.596 | Improved |
| add weighted fusion and coarse-to-fine retrieval over the hierarchy (ablation) | Anchor (Qwen2.5-72B) · Common2Sense | Average F1: 0.491 -> 0.596 | Improved |
| add LLM-elicited parameters (ablation) | Anchor (Qwen2.5-72B) · Common2Sense | Average F1: 0.483 -> 0.596 | Improved |
| add LLM-elicited parameters (ablation) | Anchor (Qwen2.5-32B) · Common2Sense | Average F1: 0.435 -> 0.572 | Improved |
| replace linear opinion pool with Bayesian model averaging (ablation) | Anchor · Common2Sense | Average F1: 0.575 vs 0.572 (32B); 0.592 vs 0.596 (72B) | No clear effect |
| use Anchor-generated factors in Bird (ablation) | Bird · Common2Sense | Average F1: 0.513 -> 0.532 | Improved |
| add Anchor hierarchical mapping to Bird with Anchor factors (ablation) | Bird · Common2Sense | Average F1: 0.532 -> 0.526 | Hurt |
| add LOP of Naive Bayes and CBN (ablation) | Bird with Anchor factors and mapping · Common2Sense | Average F1: 0.526 -> 0.568 | Improved |
Closest papers by meaning
- BIRD: A Trustworthy Bayesian Inference Framework for Large Language Models (ICLR 2025 · similarity 0.90)
- Large Language Bayes (NeurIPS 2025 · similarity 0.88)
- Anchored Diffusion Language Model (NeurIPS 2025 · similarity 0.88)
- OptScale: Probabilistic Optimality for Inference-time Scaling (AAAI 2026 · similarity 0.86)
- metabench - A Sparse Benchmark of Reasoning and Knowledge in Large Language Models (ICLR 2025 · similarity 0.86)
- Large Language Models to Enhance Bayesian Optimization (ICLR 2024 · similarity 0.85)
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.
