Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
2 results for “contrastive decoding”
SPIN Processed News Frame: The Hype
Attention-Guided Layer Selection for Contrastive Decoding in Large Language Models
Researchers introduced three new attention-guided layer selection strategies for contrastive decoding in LLMs to improve factuality, outperforming DoLa on TruthfulQA multi-answer metrics.
Spin 40% Claim Present in Source AI Risk Moderate
arXiv Computation and Language
Jul 28, 2026
SPIN Processed News Frame: The Hype
Knowledge Injection Exists in MoE? Exploring Expert-Aware Contrast Decoding in MoE for Mitigating LLMs'Hallucinations
Researchers propose EAACD, a new contrastive decoding method tailored for mixture-of-experts (MoE) LLMs that leverages expert activation differences in higher layers to reduce hallucinations on QA tasks, outperforming baselines across four datasets.
Spin 65% Claim Present in Source AI Risk Moderate
arXiv Computation and Language
Jul 24, 2026