The paper introduces DescaPE, a decoding framework that uses internal model signals to reduce hallucinations in large language models. By identifying a factual‑salient layer span and training a lightweight probe to approximate its signal, DescaPE penalizes high‑risk continuations and rewards factually grounded ones during inference. Experiments on five factuality benchmarks across three LLMs show that DescaPE improves factuality with only a 1.10× latency overhead.
By Hayeong Ryu, JungMin Yun, Byeonggeuk Lim, Sunhee Jo, YoungBin Kim
arXiv:2603. 18482v2 Announce Type: replace-cross Abstract: Standard decoding strategies for text generation, including top-$k$, nucleus sampling, and contrastive search, select tokens based on likelihood, restricting outputs to high-probability regions.
By Esteban Garces Arias, Nurzhan Sapargali, Christian Heumann, Matthias A{\ss}enmacher
arXiv:2610.23290v1 Announce Type: new
Abstract: Diffusion-based multimodal large language models (dMLLMs) decode by iterative denoising with bidirectional attention, which precludes the standard auto...
By Qiyan Zhao, Xiaofeng Zhang, Xu-Yao Zhang, Shijie Han, Zhepeng Li, Shunzhi Zhu, Da-Han Wang
SimLens is a training‑free decoder that improves early‑layer predictions in large language models by keeping only the start token and a candidate answer token and performing a lightweight continuation through the remaining layers. It outperforms direct linear readouts, yielding higher accuracy on tasks such as ARC, BoolQ, and HeadQA with LLaMA‑7B and Vicuna‑7B. The method is extended to Linear SimLens for confidence estimation and combined into SimExit, a hybrid early‑exit mechanism that achieves significant speedups while maintaining accuracy.
By Ming Ma, Bowen Zheng, Zhongqiao Lin, Tianming Yang
arXiv:2609.00796v1 Announce Type: new
Abstract: As one of the most critical challenges in large language models, contextual faithfulness directly determines their reliability in knowledge-intensive a...
By Guanqiao Chen, Di Wang, Lijie Hu
Elastic Threshold Attention (ETA) is a trainable attention mechanism that dynamically predicts contextual thresholds from query representations, enabling selective pruning of KV cache tokens during long‑context decoding. By multiplicatively suppressing sub‑threshold logits during training, ETA avoids representation collapse and eliminates localized attention sinks, allowing a 1.45B model to match dense attention performance at roughly 85% training sparsity and 38% active decode density. At inference, a custom Triton kernel achieves up to 2.5× faster decoding on sequences up to 512K tokens, and an offline calibration step can further reduce compute by 27% by freezing per‑head thresholds.
By Themistoklis Haris, Henry Li, Maryam Karimzadehgan