AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Self-Guidance: Enhancing Neural Codecs via Decoder Manifold Alignment

Xiang Li, Yixuan Zhou, Jingran Xie, Zhiyong Wu, Hui Wang

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Efficient language transformers · 0 citations (OpenAlex)

Read the paper

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 changedWhereMeasuredEffect
add self-guidance feature mapping loss (ablation)XCodec2 (65536) · LibriSpeech test-cleanPESQ: 2.28 -> 2.39Improved
add self-guidance feature mapping loss (ablation)XCodec2 (8192) · LibriSpeech test-cleanPESQ: 2.03 -> 2.13Improved
add self-guidance feature mapping loss (ablation)XCodec2 (65536) · LibriSpeech test-cleanWER: 3.2 -> 3.2No clear effect
add self-guidance feature mapping loss (ablation)XCodec2 (16384) · LibriSpeech test-cleanWER: 3.47 -> 3.53Hurt
add self-guidance to BigCodec decoder (ablation)BigCodec 40Hz · LibriSpeech test-cleanPESQ-WB: 1.6740 -> 1.7650Improved
add self-guidance to SimVQ quantizer codec (ablation)XCodec2 SimVQ · LibriSpeech test-cleanPESQ: 2.10 -> 2.17Improved
add self-guidance to ResidualFSQ codec (ablation)XCodec2 ResidualFSQ · LibriSpeech test-cleanPESQ: 1.75 -> 1.86Improved
add self-guidance loss (ablation)XCodec2 decoder · Hidden feature alignmentProcrustes residuals: 0.265 -> 0.171Improved
add self-guidance loss (ablation)XCodec2 · Quantization error mean (65536)error mean: 0.858 -> 0.851No clear effect
add self-guidance loss (ablation)XCodec2 TTS (16384) · LibriTTS test-clean TTSWER: 28.78 -> 28.02Improved
add self-guidance loss (ablation)XCodec2 TTS (65536) · LibriTTS test-clean TTSWER: 33.03 -> 35.07Hurt
reduce codebook size from 65536 to 16384 (ablation)XCodec2 TTS · LibriTTS test-clean TTSWER: 33.03 -> 28.78Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy