Large models (LMs) are powerful content generators, yet their open-ended nature can also introduce potential risks, such as generating harmful or biased content. Existing guardrails mostly perform post-hoc detection that may expose unsafe content before it is caught, and the latency constraints further push them toward lightweight models, limiting detection accuracy. In this work, we propose Kelp, a novel plug-in framework that enables streaming risk detection within the LM generation pipeline. Kelp leverages intermediate LM hidden states through a Streaming Latent Dynamics Head (SLD), which models the temporal evolution of risk across the generated sequence for more accurate real-time risk detection. To ensure reliable streaming moderation in real applications, we introduce an Anchored Temporal Consistency (ATC) loss to enforce monotonic harm predictions by embedding a benign-then-harmful temporal prior. Besides, for a rigorous evaluation of streaming guardrails, we also present StreamGuardBench-a model-grounded benchmark featuring on-the-fly responses from each protected model, reflecting real-world streaming scenarios in both text and vision-language tasks. Across diverse models and datasets, Kelp consistently outperforms state-of-the-art post-hoc guardrails and prior plug-in probes (15.61% higher average F1), while using only 20M parameters and adding less than 0.5 ms of per-token latency.

Large Language Models (LLMs) have shown unprecedented performance in various real-world applications. However, they are known to generate factually inaccurate outputs, a.k.a. the hallucination problem. In recent years, incorporating external knowledge extracted from Knowledge Graphs (KGs) has become a promising strategy to improve the factual accuracy of LLM-generated outputs. Nevertheless, most existing explorations rely on LLMs themselves to perform KG knowledge extraction, which is highly inflexible as LLMs can only provide binary judgment on whether a certain knowledge (e.g., a knowledge path in KG) should be used. In addition, LLMs tend to pick only knowledge with direct semantic relationship with the input text, while potentially useful knowledge with indirect semantics can be ignored. In this work, we propose a principled framework KELP with three stages to handle the above problems. Specifically, KELP is able to achieve finer granularity of flexible knowledge extraction by generating scores for knowledge paths with input texts via latent semantic matching. Meanwhile, knowledge paths with indirect semantic relationships with the input text can also be considered via trained encoding between the selected paths in KG and the input text. Experiments on real-world datasets validate the effectiveness of KELP.

Logic Production System (LPS) is a logic-based framework for modelling reactive behaviour. Based on abductive logic programming, it combines reactive rules with logic programs, a database and a causal theory that specifies transitions between the states of the database. This paper proposes a systematic mapping of the Kernel of this framework (called KELPS) into an answer set program (ASP). For this purpose a new variant of KELPS with finite models, called $n$-distance KELPS, is introduced. A formal definition of the mapping from this $n$-distance KELPS to ASP is given and proven sound and complete. The Answer Set Programming paradigm allows to capture additional behaviours to the basic reactivity of KELPS, in particular proactive, preemptive and prospective behaviours. These are all discussed and illustrated with examples. Then a hybrid framework is proposed that integrates KELPS and ASP, allowing to combine the strengths of both paradigms. Under consideration in Theory and Practice of Logic Programming (TPLP).

Just when we think we have a handle on all the incredible ways that dogs enhance our lives and our understanding of the world, new work with dogs expands that sphere even further. Graduate student Kiana Ehsani at the University of Washington has a great collaborator named Kelp, an Alaskan Malamute, who is a key partner in her quest to create an artificial intelligence system that thinks like a dog. The long-term goal is to produce a robot that is enough like a dog to perform many of the task that dogs are trained to do for humans. Though that may seem like a faraway dream, Ehsani's research project is edging ever closer to that possibility. Generally, the goal of the current research was to study and emulate the dog's response to visual information.

Dogs may be our best friends, but they're also our hard-working colleagues -- tasked with everything from guarding our homes to guiding visually impaired people to sniffing out bombs. And now researchers have enlisted the help of an Alaskan Malamute named Kelp to develop an artificial intelligence system that thinks just like a dog, in hopes of creating canine-like robots. To build a database of dog behavior, a team of scientists led by Kiana Ehsani, a Ph.D. student at the University of Washington, attached sensors to Kelp's paws, torso, and tail to capture her movements for a couple of hours a day while eating, playing fetch, and walking around in various indoor and outdoor environments. A camera affixed to Kelp's head recorded what she saw as she went about her everyday activities. Over the course of several weeks, the researchers amassed more than 24,000 video frames -- all associated with particular body movements.

What can artificial intelligence learn from dogs? Quite a lot, say researchers from the University of Washington and Allen Institute for AI. They recently trained neural networks to interpret and predict the behavior of canines. Their results, they say, show that animals could provide a new source of training data for AI systems -- including those used to control robots. To train AI to think like a dog, the researchers first needed data. They collected this in the form of videos and motion information captured from a single dog, a Malamute named Kelp.

In previous work, we proposed a logic-based framework in which computation is the execution of actions in an attempt to make reactive rules of the form if antecedent then consequent true in a canonical model of a logic program determined by an initial state, sequence of events, and the resulting sequence of subsequent states. In this model-theoretic semantics, reactive rules are the driving force, and logic programs play only a supporting role. In the canonical model, states, actions and other events are represented with timestamps. But in the operational semantics, for the sake of efficiency, timestamps are omitted and only the current state is maintained. State transitions are performed reactively by executing actions to make the consequents of rules true whenever the antecedents become true. This operational semantics is sound, but incomplete. It cannot make reactive rules true by preventing their antecedents from becoming true, or by proactively making their consequents true before their antecedents become true. In this paper, we characterize the notion of reactive model, and prove that the operational semantics can generate all and only such models. In order to focus on the main issues, we omit the logic programming component of the framework.