Sparse autoencoders (SAEs) are proposed to extract numerous features from large language model (LLM) representations, yet explaining these features still relies primarily on external observation. This reliance leads to superficial explanations inferred from observed model behavior and computational inefficiency from collecting such behavioral evidence at scale.
arXiv:2608. 09521v1 Announce Type: new Abstract: Activation-based tools are usually tied to one model's native hidden space, requiring probes, sparse autoencoders, and natural-language interpreters to be rebuilt or rediscovered for each new language model.
By Su-Hyeon Kim, Jiwan Mun, Yo-Sub Han
arXiv:2606. 16137v1 Announce Type: cross Abstract: Speech deepfake detection (SDD) systems require trustworthy explanations for reliable decision-making.
By Yupei Li, Qiyang Sun, Xiaoliang Wu, Chenxi Wang, Berrak Sisman, Bj\"orn W. Schuller
arXiv:2508.08966v2 Announce Type: replace
Abstract: The attention mechanism lies at the core of the transformer architecture, providing an interpretable model-internal signal that has motivated a gro...
By Marte Eggen, Jacob Lysn{\ae}s-Larsen, Inga Str\"umke
SharedSAE demonstrates that a single sparse autoencoder can replace multiple model‑specific SAEs by using a shared dictionary with model‑specific encoder‑decoder pairs. It preserves activation magnitudes, normalizes only selection scores, and supports single‑model inference via model dropout. Trained on four 1B‑scale language models, SharedSAE retains 96.6% of the mean explained variance of dedicated SAEs, shows higher cross‑model latent correlations, and allows efficient adaptation of new models to the shared latent space.
By Daniil Ognev, C\'elian Vasson, Lijie Hu, Kentaro Inui, Benjamin Heinzerling
arXiv:2602. 10352v2 Announce Type: replace-cross Abstract: Self-interpretation methods prompt language models to describe their own internal states, but remain unreliable due to hyperparameter sensitivity.
By Keenan Pepper, Alex McKenzie, Florin Pop, Stijn Servaes, Martin Leitgab, Mike Vaiana, Judd Rosenblatt, Michael S. A. Graziano, Diogo de Lucena
The paper introduces Retrieval-Augmented Decoding (RAD), a decoding-time method that improves the truthfulness of large language models without retraining. RAD uses a small reference set of up to ten annotated examples to build a grounding space of context embeddings and next-token logits, which it retrieves and aggregates during inference to shape the model’s output. Experiments on four open-ended generation benchmarks and four different LLMs show that RAD consistently outperforms strong baselines and generalizes well across tasks.
By Manh Nguyen, Sunil Gupta, Hung Le
arXiv:2608. 19529v1 Announce Type: cross Abstract: Many real-world AI systems represent entities, behaviors, and structured information using discrete machine-native symbols rather than natural language.
By Su Yan, Rakesh Iyer
arXiv:2608. 13570v1 Announce Type: cross Abstract: Latent reasoning has emerged as a powerful alternative to text-based Chain-of-Thought (CoT), offering significant gains in computational efficiency by compressing verbose reasoning into compact embeddings.
By Dayuan Zhao, Shengcao Cao, Yu-Xiong Wang, Liang-Yan Gui
The paper introduces Lens, a training‑free framework that aligns multimodal representations with the semantic perspective required by downstream tasks. Lens uses a task‑specific readout phrase to anchor the perspective and then aggregates token states after the full input, ensuring the extracted representation reflects task‑conditioned evidence integration rather than generic salient content. The method achieves a Precision@1 of 63.9 across 36 MMEB datasets, outperforming the nearest training‑free baseline by 10.2 points.
By Xinran Liu, Shouqian Shi, Yixian Chen, Ruizhi Chen, Xin-Wei Yao, Sheng Zhong
arXiv:2608.21664v1 Announce Type: new
Abstract: Safe deployment of increasingly capable models will likely come to rely on latent-space monitoring as a complement to behavioral evaluations, especiall...
By Marek Mateusz Kowalski, Joshua Fonseca Rivera, Uzay Macar, David Demitri Africa
arXiv:2607. 00684v1 Announce Type: new Abstract: The classification accuracy of pretrained Vision-Language Models (VLMs) relies on the quality of the text prompts.
By Seokhee Jin, Changhwan Sung, Sunung Mun, Hoyoung Kim, Jungseul Ok