Research map / Architectures and efficiency / Attention and transformers
Transcormer: Transformer for Sentence Scoring with Sliding Language Modeling
Kaitao Song, Yichong Leng, Xu Tan, Yicheng Zou, Tao Qin, Dongsheng Li
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 |
|---|---|---|---|
| Use bidirectional sliding language modeling with triple-stream attention instead of causal language modeling | Transcormer sentence scoring · IWSLT14 and WMT14 reranking | BLEU: 35.14 -> 35.24 (IWSLT14 De-En); 33.07 -> 33.10 (WMT14 De-En) | Improved |
| Use bidirectional sliding language modeling for rescoring | Transcormer sentence scoring · LibriSpeech reranking | WER: 2.30 -> 2.23 (dev-clean); 5.65 -> 5.54 (dev-other); 2.59 -> 2.49 (test-clean); 5.90 -> 5.72 (test-other) | Improved |
| Use SLM instead of BERT MLM for linguistic acceptability scoring | Transcormer · BLiMP Overall | accuracy: 84.2 -> 84.6 | No clear effect |
| Use SLM instead of RoBERTa large | Transcormer · BLiMP Overall | accuracy: 86.5 -> 84.6 | Hurt |
| Use the small configuration instead of the base configuration (ablation) | Transcormer · LibriSpeech dev-clean | WER: 2.23 -> 2.48 | Hurt |
| Mask more tokens per MLM pass (k=2) (ablation) | MLM rescoring · LibriSpeech dev-clean | WER: 2.30 -> 2.41 | Hurt |
| Mask more tokens per MLM pass (k=3) (ablation) | MLM rescoring · LibriSpeech dev-clean | WER: 2.30 -> 2.60 | Hurt |
| Mask n/3 tokens per MLM pass (ablation) | MLM rescoring · LibriSpeech dev-clean | WER: 2.30 -> 2.75 | Hurt |
| Mask n/2 tokens per MLM pass (ablation) | MLM rescoring · LibriSpeech dev-clean | WER: 2.30 -> 2.80 | Hurt |
| Use stream-level instead of sentence-level pre-training (ablation) | Transcormer · IWSLT14 De-En | BLEU: 35.24 -> 34.84 | Hurt |
| Use stream-level instead of sentence-level pre-training (ablation) | Transcormer · LibriSpeech dev-clean | WER: 2.23 -> 2.56 | Hurt |
| Add in-domain pre-training data (ablation) | Transcormer · IWSLT14 and WMT14 reranking | BLEU: 35.24 -> 35.74 (IWSLT14 De-En); 33.10 -> 33.51 (WMT14 De-En) | Improved |
Closest papers by meaning
- Language Generation with Strictly Proper Scoring Rules (ICML 2024 · similarity 0.84)
- Diversifying Query: Region-Guided Transformer for Temporal Sentence Grounding (AAAI 2025 · similarity 0.84)
- SparseLLM: Towards Global Pruning of Pre-trained Language Models (NeurIPS 2024 · similarity 0.83)
- A Statistical Framework for Ranking LLM-based Chatbots (ICLR 2025 · similarity 0.83)
- Language Ranker: A Lightweight Ranking framework for LLM Decoding (NeurIPS 2025 · similarity 0.83)
- Language Ranker: A Metric for Quantifying LLM Performance Across High and Low-Resource Languages (AAAI 2025 · similarity 0.83)
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.
