Research map / Vision-language and multimodal / Visual question answering
Hashing Based Answer Selection
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 |
|---|---|---|---|
| Use HAS with binary answer representations and question-answer attention | BERT-based answer selection · insuranceQA Test1 | P@1: 75.20 -> 76.38 | Helped |
| Use HAS with binary answer representations and question-answer attention | BERT-based answer selection · insuranceQA Test2 | P@1: 73.40 -> 73.71 | Helped |
| Use HAS without external QA pretraining | Answer selection · yahooQA | P@1: 74.40 -> 73.89 | Hurt |
| Use HAS without external QA pretraining | Answer selection · yahooQA | MRR: 84.00 -> 82.10 | Hurt |
| Use HAS with binary answer representations and question-answer attention | BERT-based answer selection · wikiQA | MAP: 76.57 -> 81.01 | Helped |
| Use HAS with binary answer representations instead of BERT-attention | BERT-based answer selection · insuranceQA Test2 | P@1: 74.12 -> 73.71 | Hurt |
| Use HAS with binary answer representations instead of BERT-attention | BERT-based answer selection · yahooQA | P@1: 74.78 -> 73.89 | Hurt |
| Replace full-precision stored answer matrices with HAS binary answer matrices | BERT-attention answer selection · insuranceQA | memory cost: 14.29 GB -> 0.45 GB | Helped |
| Store reusable answer representations rather than recalculating them for each question | BERT-attention answer selection · insuranceQA | time cost per question: 4.19s -> 0.28s | Helped |
| Increase beta to 5 in tanh(beta x) (ablation) | HAS · insuranceQA and wikiQA | performance | Helped |
| Vary beta within [5, 10] (ablation) | HAS · insuranceQA and wikiQA | performance | No clear effect |
| Vary the binary-constraint coefficient delta within [1e^{-7}, 1e^{-5}] (ablation) | HAS · insuranceQA and wikiQA | performance | No clear effect |
Closest papers by meaning
- Hybrid Attentive Answer Selection in CQA With Deep Users Modelling (AAAI 2018 · similarity 0.85)
- COALA: A Neural Coverage-Based Approach for Long Answer Selection with Small Data (AAAI 2019 · similarity 0.84)
- Capturing Sentence Relations for Answer Sentence Selection with Multi-Perspective Graph Encoding (AAAI 2020 · similarity 0.83)
- Answer, Assemble, Ace: Understanding How LMs Answer Multiple Choice Questions (ICLR 2025 · similarity 0.83)
- Listening to the Wise Few: Query–Key Alignment Unlocks Latent Correct Answers in Large Language Models (NeurIPS 2026 · similarity 0.82)
- Cost-efficient Knowledge-based Question Answering with Large Language Models (NeurIPS 2024 · similarity 0.81)
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.