Research map / Large language models / Language, safety and interpretability
Relevance-Promoting Language Model for Short-Text Conversation
Xin Li, Piji Li, Wei Bi, Xiaojiang Liu, Wai Lam
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 supervised source attention (SSA) and topic inference (TI) (ablation) | Transformer language model with top-k sampling · Weibo STC test | Relevance: 1.273 -> 1.606 | Improved |
| Add supervised source attention (SSA) (ablation) | Transformer language model with top-k sampling and TI · Weibo STC test | Relevance: 1.485 -> 1.606 | Improved |
| Add supervised source attention (SSA) (ablation) | Transformer language model with top-k sampling and TI · Weibo STC test | Fluency: 1.303 -> 1.346 | Hurt |
| Add topic inference (TI) (ablation) | Transformer language model with top-k sampling and SSA · Weibo STC test | Dist-2: 0.445 -> 0.544 | Improved |
| Add topic inference (TI) (ablation) | Transformer language model with top-k sampling and SSA · Weibo STC test | Relevance: 1.503 -> 1.606 | Improved |
| Use top-k sampling instead of beam search | Full relevance-promoting model · Weibo STC test | Dist-2: 0.398 -> 0.544 | Improved |
| Use top-k sampling instead of beam search | Full relevance-promoting model · Weibo STC test | Relevance: 1.515 -> 1.606 | Improved |
| Use top-k sampling instead of beam search | TFM-S2S · Weibo STC test | Relevance: 1.475 -> 1.270 | Hurt |
| Use top-k sampling instead of beam search | TFM-S2S · Weibo STC test | Dist-2: 0.290 -> 0.507 | Improved |
| Use component-wise Sigmoid instead of Softmax in topic inference | Topic inference component · Weibo STC test | Model performance | Hurt |
| Train on the concatenated query-response sequence, including query-token likelihood | Transformer language model for STC · Weibo STC test | Relevance: 1.412 -> 1.606 | Improved |
Closest papers by meaning
- Response Enhanced Semi-supervised Dialogue Query Generation (AAAI 2024 · similarity 0.85)
- Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference (AAAI 2025 · similarity 0.85)
- Making Retrieval-Augmented Language Models Robust to Irrelevant Context (ICLR 2024 · similarity 0.85)
- RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs (NeurIPS 2024 · similarity 0.85)
- Conformal Information Pursuit for Interactively Guiding Large Language Models (NeurIPS 2025 · similarity 0.85)
- Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps (NeurIPS 2026 · 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.
