In this paper, we evaluate the capability of transformer-based language models in making inferences over uncertain text that includes uncertain rules of reasoning. We cover both Pre-trained Language Models (PLMs) and generative Large Language Models (LLMs). Our evaluation results show that both generations of language models struggle with reasoning over uncertain text. We propose a novel end-to-end fine-tuning approach, Probabilistic Constraint Training (PCT), that utilizes probabilistic logical rules as constraints in the fine-tuning phase without relying on these rules in the inference stage. To assess the effectiveness of PCT, we utilize the related corpora and, additionally, create a new and more challenging benchmark that, unlike the previous ones, uses instance-specific rules. Our study demonstrates that PCT improves the transformer-based language model's intrinsic reasoning and makes their probabilistic logical reasoning process more explicit and explainable. Furthermore, PCT equips these models to effectively handle novel situations, including higher reasoning depth, new domains, and complex probabilistic structures.

Large language models (LLMs) have made impressive progress in chemistry applications, including molecular property prediction, molecular generation, experimental protocol design, etc. However, the community lacks a dialogue-based model specifically designed for chemistry. The challenge arises from the fact that most chemical data and scientific knowledge are primarily stored in structured databases, and the direct use of these structured data compromises the model's ability to maintain coherent dialogue. To tackle this issue, we develop a novel template-based instruction construction method that transforms structured knowledge into plain dialogue, making it suitable for language model training. By leveraging this approach, we develop ChemLLM, the first large language model dedicated to chemistry, capable of performing various tasks across chemical disciplines with smooth dialogue interaction. ChemLLM beats GPT-3.5 on all three principal tasks in chemistry, i.e., name conversion, molecular caption, and reaction prediction, and surpasses GPT-4 on two of them. Remarkably, ChemLLM also shows exceptional adaptability to related mathematical and physical tasks despite being trained mainly on chemical-centric corpora. Furthermore, ChemLLM demonstrates proficiency in specialized NLP tasks within chemistry, such as literature translation and cheminformatic programming. ChemLLM opens up a new avenue for exploration within chemical studies, while our method of integrating structured chemical knowledge into dialogue systems sets a new frontier for developing LLMs across various scientific fields. Codes, Datasets, and Model weights are publicly accessible at hf.co/AI4Chem/ChemLLM-7B-Chat.

Language models influence the external world: they query APIs that read and write to web pages, generate content that shapes human behavior, and run system commands as autonomous agents. These interactions form feedback loops: LLM outputs affect the world, which in turn affect subsequent LLM outputs. In this work, we show that feedback loops can cause in-context reward hacking (ICRH), where the LLM at test-time optimizes a (potentially implicit) objective but creates negative side effects in the process. For example, consider an LLM agent posting tweets with the objective of maximizing Twitter engagement; the LLM may retrieve its previous tweets into the context window and make its subsequent tweets more controversial, increasing engagement but also toxicity. We identify and study two processes that lead to ICRH: output-refinement and policy-refinement. For these processes, evaluations on static datasets are insufficient--they miss the feedback effects and thus cannot capture the most harmful behavior. In response, we provide three recommendations for evaluation to capture more instances of ICRH. As AI development accelerates, the effects of feedback loops will proliferate, increasing the need to understand their role in shaping LLM behavior.

Soft robots have immense potential given their inherent safety and adaptability, but challenges in soft actuator forces and design constraints have limited scaling up soft robots to larger sizes. Electrothermal shape memory alloy (SMA) artificial muscles have the potential to create these large forces and high displacements, but consistently using these muscles under a well-defined model, in-situ in a soft robot, remains an open challenge. This article provides a system for maintaining the highest-possible consistent SMA forces, over long lifetimes, by combining a fatigue testing protocol with a supervisory control system for the muscles' internal temperature state. We propose a design of a soft limb with swap-able SMA muscles, and deploy the limb in a blocked-force test to quantify the relationship between the measured maximum force at different temperatures over different lifetimes. Then, by applying an invariance-based control system to maintain temperatures under our long-life limit, we demonstrate consistent high forces in a practical task over hundreds of cycles. The method we developed allows for practical implementation of SMAs in soft robots through characterizing and controlling their behavior in-situ, and provides a method to impose limits that maximize their consistent, repeatable behavior.

Neoteric works have shown that modern deep learning models can exhibit a sparse double descent phenomenon. Indeed, as the sparsity of the model increases, the test performance first worsens since the model is overfitting the training data; then, the overfitting reduces, leading to an improvement in performance, and finally, the model begins to forget critical information, resulting in underfitting. Such a behavior prevents using traditional early stop criteria. In this work, we have three key contributions. First, we propose a learning framework that avoids such a phenomenon and improves generalization. Second, we introduce an entropy measure providing more insights into the insurgence of this phenomenon and enabling the use of traditional stop criteria. Third, we provide a comprehensive quantitative analysis of contingent factors such as re-initialization methods, model width and depth, and dataset noise. The contributions are supported by empirical evidence in typical setups. Our code is available at https://github.com/VGCQ/DSD2.

Motivated by the success of Transformers when applied to sequences of discrete symbols, token-based world models (TBWMs) were recently proposed as sample-efficient methods. In TBWMs, the world model consumes agent experience as a language-like sequence of tokens, where each observation constitutes a sub-sequence. However, during imagination, the sequential token-by-token generation of next observations results in a severe bottleneck, leading to long training times, poor GPU utilization, and limited representations. To resolve this bottleneck, we devise a novel Parallel Observation Prediction (POP) mechanism. POP augments a Retentive Network (RetNet) with a novel forward mode tailored to our reinforcement learning setting. We incorporate POP in a novel TBWM agent named REM (Retentive Environment Model), showcasing a 15.4x faster imagination compared to prior TBWMs. REM attains superhuman performance on 12 out of 26 games of the Atari 100K benchmark, while training in less than 12 hours. Our code is available at \url{https://github.com/leor-c/REM}.

This study investigates the integration of machine learning (ML) and data assimilation (DA) techniques, focusing on implementing surrogate models for Geological Carbon Storage (GCS) projects while maintaining high fidelity physical results in posterior states. Initially, we evaluate the surrogate modeling capability of two distinct machine learning models, Fourier Neural Operators (FNOs) and Transformer UNet (T-UNet), in the context of CO$_2$ injection simulations within channelized reservoirs. We introduce the Surrogate-based hybrid ESMDA (SH-ESMDA), an adaptation of the traditional Ensemble Smoother with Multiple Data Assimilation (ESMDA). This method uses FNOs and T-UNet as surrogate models and has the potential to make the standard ESMDA process at least 50% faster or more, depending on the number of assimilation steps. Additionally, we introduce Surrogate-based Hybrid RML (SH-RML), a variational data assimilation approach that relies on the randomized maximum likelihood (RML) where both the FNO and the T-UNet enable the computation of gradients for the optimization of the objective function, and a high-fidelity model is employed for the computation of the posterior states. Our comparative analyses show that SH-RML offers better uncertainty quantification compared to conventional ESMDA for the case study.

To foster an immersive and natural human-robot interaction, the implementation of tactile perception and feedback becomes imperative, effectively bridging the conventional sensory gap. In this paper, we propose a dual-modal electronic skin (e-skin) that integrates magnetic tactile sensing and vibration feedback for enhanced human-robot interaction. The dual-modal tactile e-skin offers multi-functional tactile sensing and programmable haptic feedback, underpinned by a layered structure comprised of flexible magnetic films, soft silicone, a Hall sensor and actuator array, and a microcontroller unit. The e-skin captures the magnetic field changes caused by subtle deformations through Hall sensors, employing deep learning for accurate tactile perception. Simultaneously, the actuator array generates mechanical vibrations to facilitate haptic feedback, delivering diverse mechanical stimuli. Notably, the dual-modal e-skin is capable of transmitting tactile information bidirectionally, enabling object recognition and fine-weighing operations. This bidirectional tactile interaction framework will enhance the immersion and efficiency of interactions between humans and robots.

We propose that small pretrained foundational generative language models with millions of parameters can be utilized as a general learning framework for sequence-based tasks. Our proposal overcomes the computational resource, skill set, and timeline challenges associated with training neural networks and language models from scratch. Further, our approach focuses on creating small and highly specialized models that can accurately execute a challenging task of which the base model is incapable of performing. We demonstrate that 125M, 350M, and 1.3B parameter pretrained foundational language models can be instruction fine-tuned with 10,000-to-1,000,000 instruction examples to achieve near state-of-the-art results on challenging cheminformatics tasks. We also demonstrate the role of successive language model fine-tuning epochs on improved outcomes, as well as the importance of both data formatting and pretrained foundational language model selection for instruction fine-tuning success.

Recent advancements in federated learning (FL) have produced models that retain user privacy by training across multiple decentralized devices or systems holding local data samples. However, these strategies often neglect the inherent challenges of statistical heterogeneity and vulnerability to adversarial attacks, which can degrade model robustness and fairness. Personalized FL strategies offer some respite by adjusting models to fit individual client profiles, yet they tend to neglect server-side aggregation vulnerabilities. To address these issues, we propose Reinforcement Federated Learning (RFL), a novel framework that leverages deep reinforcement learning to adaptively optimize client contribution during aggregation, thereby enhancing both model robustness against malicious clients and fairness across participants under non-identically distributed settings. To achieve this goal, we propose a meticulous approach involving a Deep Deterministic Policy Gradient-based algorithm for continuous control of aggregation weights, an innovative client selection method based on model parameter distances, and a reward mechanism guided by validation set performance. Empirically, extensive experiments demonstrate that, in terms of robustness, RFL outperforms the state-of-the-art methods, while maintaining comparable levels of fairness, offering a promising solution to build resilient and fair federated systems.