Over the past few years, table interpretation tasks have made significant progress due to their importance and the introduction of new technologies and benchmarks in the field. This work experiments with a hybrid approach for detecting relationships among columns of unlabeled tabular data, using a Knowledge Graph (KG) as a reference point, a task known as CPA. This approach leverages large language models (LLMs) while employing statistical analysis to reduce the search space of potential KG relations. The main modules of this approach for reducing the search space are domain and range constraints detection, as well as relation co-appearance analysis. The experimental evaluation on two benchmark datasets provided by the SemTab challenge assesses the influence of each module and the effectiveness of different state-of-the-art LLMs at various levels of quantization. The experiments were performed, as well as at different prompting techniques. The proposed methodology, which is publicly available on github, proved to be competitive with state-of-the-art approaches on these datasets.

Generative AI offers potential for educational support, but often lacks pedagogical grounding and awareness of the student's learning context. Furthermore, researching student interactions with these tools within authentic learning environments remains challenging. To address this, we present JELAI, an open-source platform architecture designed to integrate fine-grained Learning Analytics (LA) with Large Language Model (LLM)-based tutoring directly within a Jupyter Notebook environment. JELAI employs a modular, containerized design featuring JupyterLab extensions for telemetry and chat, alongside a central mid-dleware handling LA processing and context-aware LLM prompt enrichment. This architecture enables the capture of integrated code interaction and chat data, facilitating real-time, context-sensitive AI scaffolding and research into student behaviour. We describe the system's design, implementation, and demonstrate its feasibility through system performance benchmarks and two proof-of-concept use cases illustrating its capabilities for logging multi-modal data, analysing help-seeking patterns, and supporting A/B testing of AI configurations. JELAI's primary contribution is its technical framework, providing a flexible tool for researchers and educators to develop, deploy, and study LA-informed AI tutoring within the widely used Jupyter ecosystem.

Recently, scaling test-time compute on Large Language Models (LLM) has garnered wide attention. However, there has been limited investigation of how various reasoning prompting strategies perform as scaling. In this paper, we focus on a standard and realistic scaling setting: majority voting. We systematically conduct experiments on 6 LLMs $\times$ 8 prompting strategies $\times$ 6 benchmarks. Experiment results consistently show that as the sampling time and computational overhead increase, complicated prompting strategies with superior initial performance gradually fall behind simple Chain-of-Thought. We analyze this phenomenon and provide theoretical proofs. Additionally, we propose a probabilistic method to efficiently predict scaling performance and identify the best prompting strategy under large sampling times, eliminating the need for resource-intensive inference processes in practical applications. Furthermore, we introduce two ways derived from our theoretical analysis to significantly improve the scaling performance. We hope that our research can promote to re-examine the role of complicated prompting, unleash the potential of simple prompting strategies, and provide new insights for enhancing test-time scaling performance. Code is available at https://github.com/MraDonkey/rethinking_prompting.

Pioneering token-based works such as Chameleon and Emu3 have established a foundation for multimodal unification but face challenges of high training computational overhead and limited comprehension performance due to a lack of high-level semantics. In this paper, we introduce TokLIP, a visual tokenizer that enhances comprehension by semanticizing vector-quantized (VQ) tokens and incorporating CLIP-level semantics while enabling end-to-end multimodal autoregressive training with standard VQ tokens. TokLIP integrates a low-level discrete VQ tokenizer with a ViT-based token encoder to capture high-level continuous semantics. Unlike previous approaches (e.g., VILA-U) that discretize high-level features, TokLIP disentangles training objectives for comprehension and generation, allowing the direct application of advanced VQ tokenizers without the need for tailored quantization operations. Our empirical results demonstrate that TokLIP achieves exceptional data efficiency, empowering visual tokens with high-level semantic understanding while enhancing low-level generative capacity, making it well-suited for autoregressive Transformers in both comprehension and generation tasks. The code and models are available at https://github.com/TencentARC/TokLIP.

Interpreting object-referential language and grounding objects in 3D with spatial relations and attributes is essential for robots operating alongside humans. However, this task is often challenging due to the diversity of scenes, large number of fine-grained objects, and complex free-form nature of language references. Furthermore, in the 3D domain, obtaining large amounts of natural language training data is difficult. Thus, it is important for methods to learn from little data and zero-shot generalize to new environments. To address these challenges, we propose SORT3D, an approach that utilizes rich object attributes from 2D data and merges a heuristics-based spatial reasoning toolbox with the ability of large language models (LLMs) to perform sequential reasoning. Importantly, our method does not require text-to-3D data for training and can be applied zero-shot to unseen environments. We show that SORT3D achieves state-of-the-art zero-shot performance on complex view-dependent grounding tasks on two benchmarks. We also implement the pipeline to run real-time on two autonomous vehicles and demonstrate that our approach can be used for object-goal navigation on previously unseen real-world environments. All source code for the system pipeline is publicly released at https://github.com/nzantout/SORT3D.

In response to these limitations, this work introduces a hierarchical Embodied system with an Agent-based planner, named EmbodiedAgent. EmbodiedAgent leverages a next-action prediction paradigm to establish a heterogeneous multi-robot control system. The core agent generates a single action and its corresponding arguments per inference, terminating upon receiving an end-of-planning signal, thus ensuring a controlled and concise execution process. To address the aforementioned challenges, we enhance the planner's robustness and generalizability through supervised fine-tuning. Extended from previous work MultiPlan [4], we present MultiPlan+, a large-scale dataset comprising 100 scenarios with over 18,000 tasks, enriched with a subset of impractical cases to mitigate hallucinations. Additionally, we develop an agent based on a fine-tuned language model equipped with function calling capabilities and structured memory. Specifically, robot skills, termination signals, and error signals related to impractical cases are encapsulated as tools, while planning history is organized within the structured memory. For low-level execution, we employ specialized policies trained on individual basic tasks to ensure reliable and robust performance. Furthermore, we propose a comprehensive R obot Planning A ssessment S chema ( RPAS), which moves beyond error-type diagnostics to emphasize stratified success rates assessed through both human evaluation and automated grading.

The recent advent of large language models (LLMs) have brought forth a fresh set of challenges for protecting users' data privacy. LLMs and their APIs present significant privacy concerns at inference time, which are fundamentally distinct from the well-documented risks of training data memorization [19, 50, 64, 98]. While the potential adversary in training data scenarios could be any API user, the threat during inference primarily stems from the model owner--typically the organization hosting the LLM. This inference stage poses a significant privacy risk, as prompts in natural language may include various types of sensitive information, from personally identifiable data like SSNs or credit card numbers to personal health or financial details. The ensuing privacy threat is exacerbated with the growing use of in-context learning, that involves presenting the LLM with a few training examples as part of the prompt during inference [17].

At the core of FiSHformer is a novel finite admixture model of shared heads (FiSH) that samples attention matrices from a set of global attention matrices. The number of global attention matrices is much smaller than the number of local attention matrices generated. FiSHformers directly learn these global attention matrices rather than the local ones as in other transformers, thus significantly improving the computational and memory efficiency of the model.

Desination will be a red barn on the right 1. Continued on next page 18 Use Case Example rewrite Rewrite the following text to be more light-hearted: -- {very formal text} -- chat The following is a conversation with an AI assistant.