arXiv AI October 31, 2026 By Illia Makarov

One Vector or Many? Clustered Optimal-Transport Steering under Quantization

The paper investigates clustered optimal‑transport (CHaRS) steering for language models, comparing it to the standard DiM approach. It shows that CHaRS, which routes hidden states to cluster‑based displacements, can reduce toxicity by 37‑61% more than DiM on certain models while maintaining comparable fluency. The study also demonstrates that these benefits persist under INT8 and NF4 quantization, and that a simpler constant map still offers significant toxicity reduction, leaving the exact token‑routing strategy open for further exploration.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv AI

Jul 3

Breaking Safety at the Token Boundary: How BPE Tokenization Creates Exploitable Gaps in LLM Alignment

arXiv:2607. 01239v1 Announce Type: cross Abstract: Character-level perturbations bypass safety alignment in modern LLMs despite leaving prompts human-readable.

By Tung-Ling Li, Hongliang Liu, Yuhao Wu

llms nlp reinforcement-learning fine-tuning multimodal safety

arXiv Computation and Language

Sep 1

A Hub of Short Rows Inflates Intrinsic Dimension Estimation of Token Embeddings

arXiv:2608.29702v1 Announce Type: new Abstract: A token-embedding table holds a hub of short rows near its origin, and we show that this cluster biases what nearest-neighbor intrinsic-dimension (ID)...

By Alexandre Quemy

llms rag

arXiv Computation and Language

Aug 27

AgentDiff: Meaning-Bearing Rewrites Trigger Deeper Divergence than Presentation Changes in LLM Agents

The paper introduces AgentDiff, a metric that quantifies how much LLM agents’ answers differ when inputs are altered by meaning‑bearing rewrites (paraphrases, synonym substitutions) versus presentation changes (reordering, formatting, distractors). Across 68 model–benchmark–scaffold combinations involving ten LLMs and over 1,500 questions, meaning‑bearing rewrites consistently produce a roughly 20‑percentage‑point higher inconsistency rate than presentation changes, a gap that persists across severity proxies and remains significant even outside the Qwen family. Trace analysis reveals that meaning‑bearing rewrites preserve the first action but reduce thought similarity from the second step onward, extending the divergence cascade—a phenomenon termed “stealth divergence.”

By Liyun Zhang, Jiayi Guo

llms agents benchmarks

arXiv AI

Sep 10

Quantization Amplifies Determinism, Not Bias: Scale-Dependent Behavioral Effects of Serving-Time Weight Compression

arXiv:2609.07901v1 Announce Type: new Abstract: Weight quantization largely determines the economics of serving open-weight LLMs. Its costs are usually assessed with capability benchmarks, on which 4...

By Dachi Kurtskhalia

llms efficiency benchmarks safety

arXiv Machine Learning

Sep 10

Steering Under Compression: Dose-Response, Capability Cost, and Failure Asymmetry in Quantized LLMs

arXiv:2609.06473v1 Announce Type: new Abstract: Inference-time activation steering enables behavioral control of large language models without parameter modification, while post-training quantization...

By Saurav Bhandari, Benjamin Wade

llms efficiency

arXiv Machine Learning

Oct 3

One pool, many targets: a conservation layer and what archival data can identify

arXiv:2610. 00445v1 Announce Type: new Abstract: Pairwise guide--transcript scores do not enforce conservation of a finite guide-loaded RISC pool when they are interpreted independently as occupancies.

By Zahra Khodagholi, Niloofar Yousefi