Large Language Models (LLMs) have gained significant popularity in recent years for their ability to answer questions in various fields. However, these models have a tendency to "hallucinate" their responses, making it challenging to evaluate their performance. A major challenge is determining how to assess the certainty of a model's predictions and how it correlates with accuracy. In this work, we introduce an analysis for evaluating the performance of popular open-source LLMs, as well as gpt-3.5 Turbo, on multiple choice physics questionnaires. We focus on the relationship between answer accuracy and variability in topics related to physics. Our findings suggest that most models provide accurate replies in cases where they are certain, but this is by far not a general behavior. The relationship between accuracy and uncertainty exposes a broad horizontal bell-shaped distribution. We report how the asymmetry between accuracy and uncertainty intensifies as the questions demand more logical reasoning of the LLM agent, while the same relationship remains sharp for knowledge retrieval tasks.

Deploying robots in open-world environments involves complex tasks characterized by long sequences and rich interactions, necessitating efficient transfer of robotic skills across diverse and complex scenarios. To address this challenge, we propose a skill library framework based on knowledge graphs, which endows robots with high-level skill awareness and spatial semantic understanding. The framework hierarchically organizes operational knowledge by constructing a "task graph" and a "scene graph" to represent task and scene semantic information, respectively. We introduce a "state graph" to facilitate interaction between high-level task planning and low-level scene information. Furthermore, we propose a hierarchical transfer framework for operational skills. At the task level, the framework integrates contextual learning and chain-of-thought prompting within a four-stage prompt paradigm, leveraging large language models' (LLMs) reasoning and generalization capabilities to achieve task-level subtask sequence transfer. At the motion level, an adaptive trajectory transfer method is developed using the A* algorithm and the skill library, enabling motion-level adaptive trajectory transfer. At the physical level, we introduce an adaptive contour extraction and posture perception method based on tactile perception. This method dynamically obtains high-precision contour and posture information from visual-tactile texture data and adjusts transferred skills, such as contact positions and postures, to ensure effectiveness in new environments. Experimental results validate the effectiveness of the proposed methods. Project website:https://github.com/MingchaoQi/skill_transfer

6G's AI native vision of embedding advance intelligence in the network while bringing it closer to the user requires a systematic evaluation of Generative AI (GenAI) models on edge devices. Rapidly emerging solutions based on Open RAN (ORAN) and Network-in-a-Box strongly advocate the use of low-cost, off-the-shelf components for simpler and efficient deployment, e.g., in provisioning rural connectivity. In this context, conceptual architecture, hardware testbeds and precise performance quantification of Large Language Models (LLMs) on off-the-shelf edge devices remains largely unexplored. This research investigates computationally demanding LLM inference on a single commodity Raspberry Pi serving as an edge testbed for ORAN. We investigate various LLMs, including small, medium and large models, on a Raspberry Pi 5 Cluster using a lightweight Kubernetes distribution (K3s) with modular prompting implementation. We study its feasibility and limitations by analyzing throughput, latency, accuracy and efficiency. Our findings indicate that CPU-only deployment of lightweight models, such as Yi, Phi, and Llama3, can effectively support edge applications, achieving a generation throughput of 5 to 12 tokens per second with less than 50\% CPU and RAM usage. We conclude that GenAI on the edge offers localized inference in remote or bandwidth-constrained environments in 6G networks without reliance on cloud infrastructure.

Modeling disease progression is crucial for improving the quality and efficacy of clinical diagnosis and prognosis, but it is often hindered by a lack of longitudinal medical image monitoring for individual patients. To address this challenge, we propose the first Medical Video Generation (MVG) framework that enables controlled manipulation of disease-related image and video features, allowing precise, realistic, and personalized simulations of disease progression. Our approach begins by leveraging large language models (LLMs) to recaption prompt for disease trajectory. Next, a controllable multi-round diffusion model simulates the disease progression state for each patient, creating realistic intermediate disease state sequence. Finally, a diffusion-based video transition generation model interpolates disease progression between these states. We validate our framework across three medical imaging domains: chest X-ray, fundus photography, and skin image. Our results demonstrate that MVG significantly outperforms baseline models in generating coherent and clinically plausible disease trajectories. Two user studies by veteran physicians, provide further validation and insights into the clinical utility of the generated sequences. MVG has the potential to assist healthcare providers in modeling disease trajectories, interpolating missing medical image data, and enhancing medical education through realistic, dynamic visualizations of disease progression.

Adversarial robustness of neural networks is an increasingly important area of research, combining studies on computer vision models, large language models (LLMs), and others. With the release of JPEG AI - the first standard for end-to-end neural image compression (NIC) methods - the question of its robustness has become critically significant. JPEG AI is among the first international, real-world applications of neural-network-based models to be embedded in consumer devices. However, research on NIC robustness has been limited to open-source codecs and a narrow range of attacks. This paper proposes a new methodology for measuring NIC robustness to adversarial attacks. We present the first large-scale evaluation of JPEG AI's robustness, comparing it with other NIC models. Our evaluation results and code are publicly available online (link is hidden for a blind review).

As large language models (LLMs) are increasingly applied in areas influencing societal outcomes, it is critical to understand their tendency to perpetuate and amplify biases. This study investigates whether LLMs exhibit biases in predicting human mobility -- a fundamental human behavior -- based on race and gender. Using three prominent LLMs -- GPT-4, Gemini, and Claude -- we analyzed their predictions of visitations to points of interest (POIs) for individuals, relying on prompts that included names with and without explicit demographic details. We find that LLMs frequently reflect and amplify existing societal biases. Specifically, predictions for minority groups were disproportionately skewed, with these individuals being significantly less likely to be associated with wealth-related points of interest (POIs). Gender biases were also evident, as female individuals were consistently linked to fewer career-related POIs compared to their male counterparts. These biased associations suggest that LLMs not only mirror but also exacerbate societal stereotypes, particularly in contexts involving race and gender.

We propose affine concept editing (ACE) as an approach for steering language models' behavior by intervening directly in activations. We begin with an affine decomposition of model activation vectors and show that prior methods for steering model behavior correspond to subsets of terms of this decomposition. We then provide a derivation of ACE and use it to control refusal behavior on ten different models, including Llama 3 70B. ACE combines affine subspace projection and activation addition to reliably control the model's refusal responses across prompt types. We evaluate the results using LLM-based scoring on a collection of harmful and harmless prompts. Our experiments demonstrate that ACE consistently achieves more precise control over model behavior than existing methods and generalizes to models where directional ablation via affine subspace projection alone produces incoherent outputs. Code for reproducing our results is available at https://github.com/EleutherAI/steering-llama3 .

Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, from complex reasoning to code generation [Chowdhery et al., 2024, OpenAI et al., 2024, Bubeck et al., 2023, Anil et al., 2023]. Many of these advances can be attributed to Chain-of-Thought (CoT) reasoning [Wei et al., 2024, Nye et al., 2021, Li et al., 2024], which involves breaking down complex problems into a series of intermediate steps, mirroring human-like reasoning processes. The success of CoT reasoning, particularly in domains such as mathematics, logic, and multi-step decision-making, has led researchers and developers to incorporate CoT-like features directly into model training, i.e. the FLAN family of models [Chung et al., 2022, Wei et al., 2022]. This paper introduces a new formal framework for analyzing CoT in LLMs. We provide a rigorous method grounded in information theory, to evaluate the quality of each step in a model's reasoning process, thus offering insights beyond simple accuracy metrics to identify areas for improvement.

Though large language models (LLMs) have demonstrated exceptional performance across numerous problems, their application to predictive tasks in relational databases remains largely unexplored. In this work, we address the notion that LLMs cannot yield satisfactory results on relational databases due to their interconnected tables, complex relationships, and heterogeneous data types. Using the recently introduced RelBench benchmark, we demonstrate that even a straightforward application of LLMs achieves competitive performance on these tasks. These findings establish LLMs as a promising new baseline for ML on relational databases and encourage further research in this direction.

Multi-modal Large Language models (MLLMs) are always trained on data from diverse and unreliable sources, which may contain misaligned or mislabeled text-image pairs. This frequently causes robustness issues and hallucinations, leading to performance degradation. Data valuation is an efficient way to detect and trace these misalignments. Nevertheless, existing methods are computationally expensive for MLLMs. While computationally efficient, the classical influence functions are inadequate for contrastive learning models because they were originally designed for pointwise loss. Additionally, contrastive learning involves minimizing the distance between the modalities of positive samples and maximizing the distance between the modalities of negative samples. This requires us to evaluate the influence of samples from both perspectives. To tackle these challenges, we introduce the Extended Influence Function for Contrastive Loss (ECIF), an influence function crafted for contrastive loss. ECIF considers both positive and negative samples and provides a closed-form approximation of contrastive learning models, eliminating the need for retraining. Building upon ECIF, we develop a series of algorithms for data evaluation in MLLM, misalignment detection, and misprediction trace-back tasks. Experimental results demonstrate our ECIF advances the transparency and interpretability of MLLMs by offering a more accurate assessment of data impact and model alignment compared to traditional baseline methods.