Research map / Large language models / Language, safety and interpretability
Elastic Responding Machine for Dialog Generation with Dynamically Mechanism Selecting
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the filter that dynamically selects mechanisms (ERM-25 vs MARM-4) | multi-mechanism responder · Weibo dialogue test | Top-5 acceptable ratio (%): 64.60 -> 71.60 | Helped |
| use elastic mechanism selection instead of using all 25 mechanisms (ablation) | ERM-25 · Weibo dialogue test | Bad responses (%): 51.24 -> 29.85 | Helped |
| use elastic mechanism selection instead of all 25 mechanisms (ablation) | ERM-25 · Weibo dialogue test | Total diversity: 0.3073 -> 0.5308 | Helped |
| use elastic mechanism selection instead of all 25 mechanisms (ablation) | ERM-25 · Weibo dialogue test | Good responses (%): 17.64 -> 25.39 | Helped |
| elastic selection instead of fixed top-k (ablation) | ERM-25 · Weibo dialogue test | Diversity-F1: 0.5064 -> 0.5144 | Helped |
| ERM filter on 25 mechanisms vs MARM-25 without filter | multi-mechanism responder · Weibo dialogue test | Top-5 acceptable ratio (%): 51.93 -> 71.60 | Helped |
| increase mechanism count 4 to 25 without a filter | MARM · Weibo dialogue test | Top-5 acceptable ratio (%): 64.60 -> 51.93 | Hurt |
| ERM-25 vs MARM-4 | multi-mechanism responder · Weibo dialogue test | BLEU-4 (%): 11.56 -> 12.23 | No clear effect |
| respond with more mechanisms (larger top-k) | ERM-25-All / MARM-25 · Weibo dialogue test | Diversity-Precision vs Recall | Mixed |
Closest papers by meaning
- Mechanism-Aware Neural Machine for Dialogue Response Generation (AAAI 2017 · similarity 0.90)
- DialogBERT: Discourse-Aware Response Generation via Learning to Recover and Rank Utterances (AAAI 2021 · similarity 0.85)
- A Hierarchical Latent Variable Encoder-Decoder Model for Generating Dialogues (AAAI 2017 · similarity 0.84)
- AvgOut: A Simple Output-Probability Measure to Eliminate Dull Responses (AAAI 2020 · similarity 0.84)
- Paralinguistics-Aware Speech-Empowered Large Language Models for Natural Conversation (NeurIPS 2024 · similarity 0.84)
- Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks (ICLR 2025 · similarity 0.82)
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.