Research map / Large language models / Language, safety and interpretability
CASE: Context-Aware Semantic Expansion
Jialong Han, Aixin Sun, Haisong Zhang, Chenliang Li, Shuming Shi
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 the CASE model with NBoW context encoding rather than PIC | CASE · CASE test | Recall@10: 19.62 -> 23.42 | Improved |
| Add seed-oblivious attention to NBoW (ablation) | NBoW · CASE test | Recall@10: 23.42 -> 23.88 | Improved |
| Add trans-dot seed-aware attention to NBoW (ablation) | NBoW · CASE test | Recall@10: 23.42 -> 24.51 | Improved |
| Use concat rather than trans-dot seed-aware attention (ablation) | NBoW · CASE test | Recall@10: 24.51 -> 24.12 | Hurt |
| Use dot rather than trans-dot seed-aware attention (ablation) | NBoW · CASE test | Recall@10: 24.51 -> 22.13 | Hurt |
| Remove context encoding (ablation) | CASE · CASE test | Recall@10: 23.42 -> 15.81 | Hurt |
| Replace NBoW with CNN as context encoder (ablation) | CASE · CASE test | Recall@10: 23.42 -> 20.97 | Hurt |
| Replace NBoW with LSTM as context encoder (ablation) | CASE · CASE test | Recall@10: 23.42 -> 19.02 | Hurt |
| Remove hypernyms from the context (ablation) | NBoW + trans-dot · CASE test | Recall@10: 24.51 -> 23.41 | Hurt |
| Add co-occurrence weighting to LS to form LSCo | LS · CASE test | Recall@10: 2.84 -> 3.54 | Improved |
Closest papers by meaning
- Provence: efficient and robust context pruning for retrieval-augmented generation (ICLR 2025 · similarity 0.86)
- Retrieval is Accurate Generation (ICLR 2024 · similarity 0.85)
- The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentation (NeurIPS 2025 · similarity 0.85)
- AdmTree: Compressing Lengthy Context with Adaptive Semantic Trees (NeurIPS 2025 · similarity 0.85)
- Leveraging Attention to Effectively Compress Prompts for Long-Context LLMs (AAAI 2025 · similarity 0.84)
- Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference (AAAI 2025 · similarity 0.84)
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.
