TRACE is a new benchmark that evaluates the safety of Large Reasoning Models (LRMs) across the entire inference pipeline, including prompts, reasoning traces, and final responses. It provides prompts in two languages covering nine risk categories and ten attack strategies, and for each prompt four LRMs generate traces and responses that are annotated for safety with supporting evidence extracted from the source text. Evaluation of 18 guardrail models on TRACE shows that detecting unsafe content in reasoning traces is much harder than in prompts or final responses, and that current models struggle to extract the necessary evidence.
By Zhenyu Wu, Siyuan Chen, Changchun Yang, Jiaqi Dong, Min Zhou, Ali Almadan, Talal Hammad, Faisal Wahbo, Aminullah Tora, Mona Alshahrani, Xin Gao
arXiv:2608. 15673v1 Announce Type: cross Abstract: Large language model guardrails can be viewed as policy-consistency problems: a system must determine which policy-relevant facts hold in a prompt-response pair and what those facts imply under a given policy.
By Satchit Chatterji, Shihan Wang, Giovanni Sileno, Erman Acar
arXiv:2608. 19299v1 Announce Type: new Abstract: Air traffic control (ATC) communication is a safety-critical dialogue that remains largely human-driven even as other parts of air traffic management have been semi-automated.
By Mahyar Ghazanfari, Matthias Casanova, Jordan Kam, Alex Zongo, Peng Wei, Torsten Darrell, Alexandre Bayen
arXiv:2608. 03291v1 Announce Type: cross Abstract: Chain-of-thought (CoT) reasoning improves large language model (LLM) performance while also providing an observable interface to the model's reasoning process.
By Shashwat Sourav, Aishwarya Balwani
arXiv:2603. 25450v2 Announce Type: replace Abstract: Detecting when a language model is wrong without ground truth labels is a fundamental challenge for safe deployment.
By Matt Gorbett, Suman Jana
SAGE (State‑Grounded, Abstention‑Aware Evaluation) is a new framework for assessing task‑oriented dialogue agents that focuses on whether each turn correctly advances the workflow state rather than just the quality of the reply. It compiles workflow specifications and per‑turn state differences into schema‑grounded criteria, then evaluates them through a cascade of symbolic rules and encoder/NLI verifiers that abstain instead of guessing, producing a turn‑level decision with an evidence trace. In experiments across MultiWOZ, Schema‑Guided Dialogue, and ABCD datasets, SAGE‑Core—using only symbolic rules and on‑device encoders—outperforms all evaluated LLM‑based judges, including GPT‑4.1 variants, while incurring zero paid LLM cost.
By Rayan Khoury, Shih-Yao Lin, Pratyush Mishra
arXiv:2606. 30128v1 Announce Type: new Abstract: Chain-of-thought (CoT) prompting improves LLM reasoning, but the source is contested: do the intermediate steps help because they carry useful semantic content, or because conditioning on more tokens buys extra computation before the model commits to an answer?
By Wenlong Wang, Fergal Reid
PROOF is a benchmark that profiles the reliability of object-level facts in instruction-tuned language models by converting a frozen Wikidata snapshot into 18,486 English multiple-choice questions grounded in 11,779 semantic facts across 101 classes, 392 properties, and 14 domains. Each question includes an explicit "I don't know" option, a "No correct option" control, and nine controlled formulations, with 1,849 questions designed as no-correct-option traps. The study evaluates 18 open-weight model deployments on 166,374 prompts, revealing wide variability in factual accuracy, sensitivity to wording changes, and the impact of decoder perturbations.
By Andrei Chetvergov, Mikhail Solovev, Timofei Sivoraksha, Stepan Ukolov, Valeriia Kuschenko, Alexander Evseev, Sergey Bolovtsov
arXiv:2608. 08700v1 Announce Type: new Abstract: Reliable evaluation of tool routing is critical as Large Language Models increasingly operate as autonomous agents.
By Dongjie Xu, Julius, Hanchi Dong, Minghua Tang, Yuxuan Sun, Ziwei Nie, Zicheng Liu, Dujun Qing, Jiajie Xu
arXiv:2607. 27031v1 Announce Type: new Abstract: Reports on how sparsification, compression, and lottery tickets change model behavior have been mixed in the prior literature, with beneficial effects observed in some studies and adverse effects in others.
By Bum Jun Kim
CARGO is a framework for evaluating agentic AI systems in production that addresses the problem of reference-instance divergence (RID), where reference-based judges penalize correct answers that involve different entity identifiers. It treats retrieved references as procedural exemplars, grounds judgments in the live instance’s context, assigns a three-way status to claims, and gates evaluation by retrieval confidence. Using the CARGO-Bench diagnostic suite, CARGO eliminates false penalties and improves discrimination while revealing a limitation in detecting procedural corruptions.
By Mukul Chhabra, Shail Patel, Luigi Medrano
The paper introduces a signed lexical gate that combines a sentence classifier’s logit margin with a sparse lexical model’s support for the predicted intent, assigning positive evidence to lexical agreement and negative evidence to a lexically favored competing intent. This gate retains more information than unsigned lexical confidence or a hard agreement rule and is calibrated via an independent binomial procedure to meet specified risk targets. Experiments on BANKING77, CLINC150, and HWU64 show that the proposed score reduces the area under the risk‑coverage curve by up to 15.8% and increases accepted coverage at low error rates, offering a compact, interpretable confidence enhancement for risk‑calibrated intent routing.
By Yezhou Cheng, Zehua Yang, Bojun Lin