Yang Liu

Kaiming Liu, Fuwen Luo, Ziyue Wang et al.

Collaborative dialogue can end with apparent agreement while participants still differ on goals, assumptions, or execution plans, creating an \textbf{illusion of alignment (IoA)}. A real-user study across 18 meetings confirms that IoA arises routinely in human collaboration. Yet IoA poses a paradox: if participants were aware of such disagreements, they would already be explicit; if not, they cannot articulate them when asked, leaving IoA invisible to both participants and observers. In this work, we make IoA detectable by generating diagnostic multiple-choice questions whose divergent answers across participants provide direct behavioral evidence of hidden disagreement. We construct \textbf{IoA-Suite}, a dataset and evaluation protocol for detecting hidden disagreement, spanning five task types and six domains. We find that even the best model attains only 49.5\% F1, with the bottleneck traced to private context that the dialogue does not surface. We then train \textbf{IoA-Prober-8B} based on IoA-Suite, reaching 51.8\% F1 on IoA-Suite. Across the aforementioned 18 real meetings, it surfaces 2.89 hidden disagreements per meeting that participants confirm they had not voiced, transferring to live human dialogue. Further, in multi-agent collaboration, pairing IoA-Prober-8B with LLM agents improves downstream task performance on BigCodeBench-Hard and HiddenBench.

Chen Li, Sijie Cheng, Yuelin Zhang et al.

Embodied agents using LLM-based planners often struggle with physical hallucinations, poor generalization to long-horizon tasks, and lack of environmental awareness. We propose GraphThink, a novel framework that integrates a task graph to provide structured knowledge for robust planning and a scene graph to maintain environmental memory for event-driven replanning. Specifically, the task graph guides LLM thinking through contextual prompting and iterative refinement, effectively mitigating planning hallucinations. Furthermore, within the GRPO framework, the task graph offers delicate reward design to train the LLM planner, enhancing long-horizon planning capabilities and improving generalization. Finally, an event-driven replanning module, powered by the scene graph, enables closed-loop environment awareness and error correction. GraphThink achieves state-of-the-art performance on the ALFRED benchmark. In particular, our high-level planner surpasses leading API-based LLMs on both the validation set and held-out long-horizon tasks, underscoring its robust zero-shot and few-shot capabilities. Additional evaluations further demonstrate strong out-of-distribution generalization to novel tasks and environments.

Liangdong Lu, Guanmin Guo, Yang Liu et al.

Bicycle (two-block circulant) quantum low-density parity-check (LDPC) codes include some of the best known small quantum codes, yet their design has relied on group-algebra formulations in which the dimension and distance are accessible only through matrix computation. We show that in the cyclic case the construction collapses into the polynomial ring $\F_2[x]/(x^{l}-1)$: self-orthogonality is automatic, the quantum dimension is read off from a polynomial gcd, and the minimum distance is certified exactly through the Calderbank correspondence to additive codes over $\F_4$, turning code search into an algebraically pre-filtered enumeration that reaches parameter regimes poorly covered by existing tables. A computer search based on this framework recovers the short codes $[[42,12,4]]_2$ and $[[62,12,4]]_2$ and produces a family of codes with competitive figure of merit $kd^2/n$, including $[[66,20,7]]_2$ with $kd^2/n=14.85$, above the bivariate bicycle code $[[144,12,12]]_2$ ($kd^2/n=12$) at less than half the block length, together with $[[46,2,8]]_2$, $[[66,2,9]]_2$, $[[66,4,8]]_2$, $[[66,6,8]]_2$ and, at $n=90$, $[[90,16,6]]_2$, $[[90,18,6]]_2$, $[[90,20,6]]_2$. An exhaustive census at $n=48$ delineates the boundary of this picture: we exhibit a $[[48,10,6]]_2$ code from a minimal $48$-element group (the Aydin--Tamo--Barg realization uses $72$ elements), and prove that distance $5$ forces a stabilizer-rank loss, which excludes $[[48,10,5]]_2$ from the weight-$8$ symmetric coset family. The framework thus opens a systematic route to bicycle-type quantum LDPC codes beyond the reach of group-theoretic searches, and identifies exactly where genuinely coset-theoretic phenomena begin.

Hao Sui, Simeng Qin, Jie Liao et al.

Agent skills, bundles of instructions and resources that an LLM agent loads on demand, form an emerging supply chain where a single poisoned skill can persistently compromise every agent that installs it. However, existing skill attacks either fire on every request or rely on fine-tuned weights or multiple skills, leaving a conditional and low-cost backdoor unexplored. In this work, we present ElasticBack, an effective conditional single-skill backdoor that plants a rule R in the skill document and a benign-looking trigger T in the user query, so the malicious payload fires only when both co-occur. ElasticBack binds the two sides through a trigger-as-switch construction, generating R via semantic-anchored rule injection. It then freezes R and evolves T against it with a stealth-constrained genetic search, so that effectiveness and stealth are optimized, keeping the backdoor weight-free and dormant on benign inputs. Extensive experiments across three target behaviors (50 skills each) and four agent LLMs show that ElasticBack attains a high attack success rate at a near-zero false-positive rate with preserved clean accuracy, transfers across models, and evades deployment-time defenses. These results motivate stronger defenses for the skill supply chain.

Yang Liu, Shiwei Hou, Xiyuan Chen et al.

EDA scripting with tool-specific, often undocumented APIs remains a long-tail bottleneck that existing LLMs fail to address. This paper presents ZhuLong, an execution-grounded LLM coding agent for PyAether and SKILL that combines API retrieval, documentation inspection, and sandbox execution via unified MCP tools, augmented by an offline API self-exploration mechanism that infers undocumented API behaviors through counterfactual experimentation. We evaluate ZhuLong on EDA-Eval-PyAether, a benchmark of 158 real-world tasks with assertion-based execution, where the complete system achieves 78.5% Pass@1 in the commercial Empyrean Aether environment, substantially outperforming a pure LLM baseline (23.6%). Ablation studies identify sandbox execution as the dominant performance driver (41.2 pp drop when removed), with the self-exploration mechanism contributing an additional 3.2 pp accuracy gain and a 22.1% reduction in per-task tool calls. On 20 interactive tasks involving unsaved layouts and schematics, ZhuLong achieves 60.0% Pass@1 for PyAether and 50.0% for SKILL.