Brownian stochastic interpolants connect paired endpoint samples, yet their sampling error can depend on displacement moments even when neither endpoint has a first moment. For arbitrary Borel couplings, we prove early-stopped, target-to-output KL guarantees for standard Euler sampling from the true source. A fourth displacement moment gives an explicit $O(N^{-1})$ bound; a second moment suffices for fixed-law convergence as drift error vanishes. For one fixed coupling with no displacement moment above two, original bridge-path KL stays of order one, comparison-path KL is of order $1/\log N$, and terminal KL is at most $C/N^2$ on the same grid. The path decomposition identifies the information about the source discarded by Markov projection. When the original bridge is Markov, current-state Euler sampling instead has a second-moment $O((M_2+d)/N)$ bound; finite-entropy Schroedinger bridges for the probability Brownian reference are included. We also construct a current-state learner with expected output KL consistency from iid endpoint pairs. Numerical diagnostics illustrate the distinct errors. Conditional diffusion identities specify a restoration interface, while image controls find no iterative advantage over cheap probabilistic predictors. Uniform terminal rates for general second-moment couplings remain open.
We study online interval selection with binary predictions of membership in a final optimal solution. Requests arrive in arbitrary order; accepted requests may be revoked, but rejected or revoked requests cannot be recovered. With rewards proportional to interval length, the best consistency under finite strict robustness is $d=(3+\sqrt5)/2$ when both guarantees hold almost surely on each fixed input, including for randomized algorithms. A continuous lower bound forces consistency strictly above $d$ at every robustness budget below $4\varphi+3\approx9.472$, whereas our deterministic algorithm attains $d$ with exact robustness $\varphi^5\approx11.090$, where $\varphi=(1+\sqrt5)/2$. Each held interval inherits a ledger recording positive prediction mass lost through irrevocable decisions. A tunable replacement cost gives a continuous family with exact consistency throughout and exact robustness on two parameter intervals, including $(C,R)=(25/9,10)$ and $(4,7)$. Randomization permits expected consistency approaching one with finite expected robustness, whereas almost-sure robustness forces expected consistency strictly above two at every finite budget.
Guan Wang University of the Chinese Academy of Sciences and Institute of Information Engineering, Chinese Academy of Sciences, Biyu Zhou, Xuehai Tang, Jizhong Han Chinese Academy of Science, Songlin Hu
To address the increasingly severe safety risk of large language models (LLMs), reasoning-based safety alignment methods have emerged. These methods overcome the limitations of ’shallow alignment’ by exposing the model’s Chain-of-Thought (CoT), enabling auditability of safety reasoning process through both training-phase supervision and post-generation verification. However, this transparency creates a critical vulnerability, a tension we define as the Security Auditability Dilemma: while explicit reasoning is a prerequisite for safety, its textual Auditable paradoxically transforms it into an optimization target for adaptive attackers and induces the model to unintentionally copy harmful content from its own reasoning context. To address this, we propose Auditable Latent CoT Alignment (ALCA), a framework that decouples internal reasoning from external output. ALCA shifts the safety deliberation process into a continuous latent space. This allows the safety reasoning process to guide the generation of harmless outputs, while eliminates the discrete textual surface that facilitates internal copying and adaptive attack. Yet, this process is not a black box. we introduce a restricted Self-Decoding mechanism that allows the model to reconstruct its latent reasoning into human-readable text for supervision under specific guidance. Extensive experiments show that ALCA achieves robustness alignment, reducing the success rate of adaptive jailbreak attacks by over 40% compared to strong baselines, while preserving performance. Our framework presents a path toward building LLMs that are both robustly secure and auditable.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Guan Wang University of the Chinese Academy of Sciences and Institute of Information Engineering, Chinese Academy of Sciences, Xuehai Tang, Biyu Zhou, Jizhong Han Chinese Academy of Science, Songlin Hu
Prevailing safety alignment methods still leave Large Language Models (LLMs) vulnerable to sophisticated jailbreak attacks. To bolster defenses, explicit reasoning mechanisms like Safety-oriented Chain-of-Thought (SCoT) have emerged, significantly enhancing robustness. However, this transparency introduces a critical trade-off: the exposed reasoning process itself becomes a new attack surface, risking the leakage of harmful information and revealing the model’s safety logic to adversaries. This paper directly confronts this dilemma, asking: Can we achieve the full benefits of deliberative safety without the costs of explicit reasoning generation? We propose Safety Reasoning Internalization to make the deliberative process in SCoT “available but not visible”. This approach is grounded in a key theoretical insight: the corrective influence of an SCoT can be effectively approximated by a targeted, low-rank update to the model’s Feed-Forward Network (FFN) layers. We operationalize this through Hierarchical Internalization of Adversarially-Guided Reasoning (HIAR), a layer-wise safety alignment framework that internalizes safety reasoning into an implicit computational pathway using Low-Rank Adaptation (LoRA). HIAR enables the model to reach a safe conclusion within a single forward pass, entirely eliminating the need to generate vulnerable SCoT text. Extensive experiments on various LLMs demonstrate that HIAR achieves a 43% lower Attack Success Rate (ASR) against distinct jailbreak attacks compared to strong baselines.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表