Research map / Large language models / Retrieval and knowledge
When Do Hallucinations Arise? A Graph Perspective on the Evolution of Path Reuse and Path Compression
Xinnan Dai, Kai Yang, Cheng Luo, Shenglai Zeng, Kai Guo, Jiliang Tang
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 |
|---|---|---|---|
| train longer on the full path search space (overfitting stage) | 6-layer LLaMA-2 from scratch · SBM graph, 10 communities, pin=0.6 | accuracy: 1.0 -> 0.98 | Hurt |
| increase number of communities from 10 to 50 (ablation) | 6-layer LLaMA-2 from scratch · SBM graph 1000 nodes | accuracy degradation: 0.97 -> 0.85 | Hurt |
| train longer with pin=0.1, pout=0.02 | 6-layer LLaMA-2 from scratch · SBM graph 1000 nodes, 1000 communities | accuracy: 0.95 -> 0.80 | Hurt |
| increase number of transformer layers (ablation) | decoder-only Transformer · path compression, uncompressed ratio | uncompressed ratio | No clear effect |
| use shallower model (ablation) | decoder-only Transformer · path compression, uncompressed ratio | uncompressed ratio | Improved |
| switch architecture among LLaMA-2, Qwen-2, Mixtral (ablation) | decoder-only Transformer · path compression, uncompressed ratio | uncompressed ratio | No clear effect |
| switch next-token prediction framework (ablation) | decoder-only Transformer · path compression | uncompressed ratio | No clear effect |
| use PPO instead of SFT for finetuning recovery (ablation) | path-compressed pretrained Transformer · SBM graph 1000 nodes, 10 communities | max accuracy improvement: 0.2 -> 0.3 | Improved |
| SFT finetuning on strongly path-compressed model (epoch 6) (ablation) | pretrained Transformer · SBM graph 1000 nodes | accuracy | Hurt |
| finetune on non-overlapping out-of-domain data (ablation) | SFT · SBM graph 1000 nodes | accuracy | Mixed |
| increase training data fraction for intrinsic reasoning (0.1% vs 20% train set) (ablation) | Transformer on ER graph · ER graph N=10 intrinsic reasoning | hallucination persistence | Hurt |
| train language-sequence models longer after loss converges | Qwen3-style Transformer, 6 to 24 layers · story language sequences | uncompressed path ratio | Hurt |
Closest papers by meaning
- Auditing Meta-Cognitive Hallucinations in Reasoning Large Language Models (NeurIPS 2025 · similarity 0.91)
- Mitigating Hallucinations in Large Language Models via Causal Reasoning (AAAI 2026 · similarity 0.91)
- Tracing the Cascade: A Topology-Aware Evaluation Framework for Scientific Agent Hallucinations (NeurIPS 2026 · similarity 0.90)
- Joint Evaluation of Answer and Reasoning Consistency for Hallucination Detection in Large Reasoning Models (AAAI 2026 · similarity 0.89)
- Thinking in Uncertainty: Mitigating Hallucinations in MLRMs with Latent Entropy-Aware Decoding (CVPR 2026 · similarity 0.89)
- More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Models (NeurIPS 2025 · similarity 0.89)
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.
