A benchmark audit with an intentionally simple graph heuristic suggests that strong performance on standard benchmarks does not always demonstrate advanced sequential, semantic, or generative modeling ability, and calls for more careful dataset selection and dataset-level diagnostic analysis when using benchmarks to su...
Hao-Yu Han, Li Ma, Han-Bing Wang et al.· arXiv.org· 1 citation
LLM-based agents are increasingly used to complete complex, long-horizon tasks without human intervention. To make progress toward the task goal, agents need to perform different duties, such as planning and verification, when needed. To understand how single agents manage these duties during long-horizon execution, we...
Xue-Xin Shen, Lingyong Yan, Shuai-Qiang Wang et al.· 0 citations
LLM agents retrieve untrusted content and act through tools, creating indirect prompt-injection risks that can cause unauthorized actions or persistent state changes. Existing automated red-teaming largely optimizes payloads for pre-specified scenarios, leaving latent vulnerabilities across the agent's behavior space u...
Peng-Fei He, Deepanjan Mitra, Vishesh Sharma et al.· 0 citations
Knowledge tracing (KT) aims to predict learners’ mastery levels of knowledge components (KCs) or test items based on their interaction records with educational content. Despite significant advancements in KT models, such as RNN-based sequence models and Transformer-based attention models, a critical limitation persists...
Ming-Liang Hou, Zi-Tao Liu, Ren-Qiang Luo et al.· ACM Transactions on Informat...· 1 citation
ConnectionMind is presented, a production-ready recommendation framework that tightly integrates the social network structure with large language models (LLMs) to enable scalable, interpretable, and reasoning-aware personalization in Meta.
Haoyu Han, Yuming Liu, Lei Huang et al.· ACM International Conference...· 0 citations
This work introduces BONSAI : Branching-Optimized Node Structure for Adaptive Identifiers, a novel framework that co-designs textual term IDs and their underlying decoding trie that satisfies two desirable properties for a highly performant trie.
This work comprehensively investigates computation-efficient strategies to speed up latent adversarial training from two complementary perspectives, and reduces per-step adversarial-training FLOPs by 48.1% while requiring only 0.0118% trainable parameters.