A rubber duck that was stuck to a seaside rock for more than a year has proved the strength of a new sticky material. The adhesive could be used in deep-sea robots and repair work, or as surgical glue for medical procedures. "We developed a super-adhesive hydrogel that works extremely well even underwater – something very few materials can achieve," says Hailong Fan at Shenzhen University in China. Hydrogels are stretchy and soft materials. Fan, then at Hokkaido University in Japan, and his colleagues analysed 24,000 sticky protein sequences from many different organisms to identify the stickiest combinations of amino acids, the building blocks of proteins.

The 2024 election cycle saw artificial intelligence deployed by political campaigns for the very first time. While candidates largely avoided major mishaps, the tech was used with little guidance or restraint. Now, the National Democratic Training Committee (NDTC) is rolling out the first official playbook making the case that Democratic campaigns can use AI responsibly ahead of the midterms. In a new online training, the committee has laid out a plan for Democratic candidates to leverage AI to create social content, write voter outreach messages, and research their districts and opponents. Since NDTC's founding in 2016, the organization says, it has trained more than 120,000 Democrats seeking political office.

We study the optimization and sample complexity of gradient-based training of a two-layer neural network with quadratic activation function in the high-dimensional regime, where the data is generated as $y \propto \sum_{j=1}^{r}λ_j σ\left(\langle \boldsymbol{θ_j}, \boldsymbol{x}\rangle\right), \boldsymbol{x} \sim N(0,\boldsymbol{I}_d)$, $σ$ is the 2nd Hermite polynomial, and $\lbrace\boldsymbolθ_j \rbrace_{j=1}^{r} \subset \mathbb{R}^d$ are orthonormal signal directions. We consider the extensive-width regime $r \asymp d^β$ for $β\in [0, 1)$, and assume a power-law decay on the (non-negative) second-layer coefficients $λ_j\asymp j^{-α}$ for $α\geq 0$. We present a sharp analysis of the SGD dynamics in the feature learning regime, for both the population limit and the finite-sample (online) discretization, and derive scaling laws for the prediction risk that highlight the power-law dependencies on the optimization time, sample size, and model width. Our analysis combines a precise characterization of the associated matrix Riccati differential equation with novel matrix monotonicity arguments to establish convergence guarantees for the infinite-dimensional effective dynamics.

This paper introduces a parameterization framework for controlling conversation quality in large language models. We explore nine key parameters across six dimensions that enable precise specification of dialogue properties. Through experiments with state-of-the-art LLMs, we demonstrate that parameter-based control produces statistically significant differences in generated conversation properties. Our approach addresses challenges in conversation generation, including topic coherence, knowledge progression, character consistency, and control granularity. The framework provides a standardized method for conversation quality control with applications in education, therapy, customer service, and entertainment. Future work will focus on implementing additional parameters through architectural modifications and developing benchmark datasets for evaluation.

The marking problem consists in assessing typed student answers for correctness with respect to a ground truth solution. This is a challenging problem that we seek to tackle using a combination of a computer algebra system, an SMT solver and a term rewriting system. A Large Language Model is used to interpret and remove errors from student responses and rewrite these in a machine readable format. Once formalized and language-aligned, the next step then consists in applying automated reasoning techniques for assessing student solution correctness. We consider two methods of automated theorem proving: off-the-shelf SMT solving and term rewriting systems tailored for physics problems involving trigonometric expressions. The development of the term rewrite system and establishing termination and confluence properties was not trivial, and we describe it in some detail in the paper. We evaluate our system on a rich pool of over 1500 real-world student exam responses from the 2023 Australian Physics Olympiad.

Recent advancements in artificial intelligence (AI), particularly in large language models (LLMs) such as OpenAI-o1 and DeepSeek-R1, have demonstrated remarkable capabilities in complex domains such as logical reasoning and experimental coding. Motivated by these advancements, numerous studies have explored the application of AI in the innovation process, particularly in the context of scientific research. These AI technologies primarily aim to develop systems that can autonomously conduct research processes across a wide range of scientific disciplines. Despite these significant strides, a comprehensive survey on AI for Research (AI4Research) remains absent, which hampers our understanding and impedes further development in this field. To address this gap, we present a comprehensive survey and offer a unified perspective on AI4Research. Specifically, the main contributions of our work are as follows: (1) Systematic taxonomy: We first introduce a systematic taxonomy to classify five mainstream tasks in AI4Research. (2) New frontiers: Then, we identify key research gaps and highlight promising future directions, focusing on the rigor and scalability of automated experiments, as well as the societal impact. (3) Abundant applications and resources: Finally, we compile a wealth of resources, including relevant multidisciplinary applications, data corpora, and tools. We hope our work will provide the research community with quick access to these resources and stimulate innovative breakthroughs in AI4Research.

Large language models (LLMs) possess strong multilingual capabilities, and combining Reinforcement Learning from Human Feedback (RLHF) with translation tasks has shown great potential. However, we observe that this paradigm performs unexpectedly poorly when applied to colloquial subtitle translation tasks. In this work, we investigate this issue and find that the offline reward model (RM) gradually diverges from the online LLM due to distributional shift, ultimately leading to undesirable training outcomes. To address this, we propose RIVAL, an adversarial training framework that formulates the process as a min-max game between the RM and the LLM. RIVAL iteratively updates the both models, with the RM trained to distinguish strong from weak translations (qualitative preference reward), and the LLM trained to enhance its translation for closing this gap. To stabilize training and improve generalizability, we also incorporate quantitative preference reward (e.g., BLEU) into the RM, enabling reference-free quality modeling aligned with human evaluation. Through extensive experiments, we demonstrate that the proposed adversarial training framework significantly improves upon translation baselines.

Large language models (LLMs) are increasingly being adopted in educational settings. These applications expand beyond English, though current LLMs remain primarily English-centric. In this work, we ascertain if their use in education settings in non-English languages is warranted. We evaluated the performance of popular LLMs on four educational tasks: identifying student misconceptions, providing targeted feedback, interactive tutoring, and grading translations in eight languages (Mandarin, Hindi, Arabic, German, Farsi, Telugu, Ukrainian, Czech) in addition to English. We find that the performance on these tasks somewhat corresponds to the amount of language represented in training data, with lower-resource languages having poorer task performance. Although the models perform reasonably well in most languages, the frequent performance drop from English is significant. Thus, we recommend that practitioners first verify that the LLM works well in the target language for their educational task before deployment.

Imagine a robot controller with the ability to adapt like human synapses, dynamically rewiring itself to overcome unforeseen challenges in real time. This paper proposes a novel zero-shot adaptation mechanism for evolutionary robotics, merging a standard Genetic Algorithm (GA) controller with online Hebbian plasticity. Inspired by biological systems, the method separates learning and memory, with the genotype acting as memory and Hebbian updates handling learning. In our approach, the fitness function is leveraged as a live scaling factor for Hebbian learning, enabling the robot's neural controller to adjust synaptic weights on-the-fly without additional training. This adds a dynamic adaptive layer that activates only during runtime to handle unexpected environmental changes. After the task, the robot 'forgets' the temporary adjustments and reverts to the original weights, preserving core knowledge. We validate this hybrid GA-Hebbian controller on an e-puck robot in a T-maze navigation task with changing light conditions and obstacles.

Strong and persistent vibration is harmful for both human and machine health. In humans, long exposure to vibrating power tools may induce health problems, such as the hand-arm vibration syndrome [1]. Instead in machines, vibration undermines the precision in control applications and may lead to mechanical wear [2, 3]. For these reasons, vibration suppression is an important capability for employing collaborative robots in new environments such as construction sites [4] where the vibration is a common phenomenon. This work builds on our preliminary results [5], in which we studied the feedforward vibration suppression in human-robot collaboration (HRC). The main outcome of our study was that feedforward force control increases the vibration suppression performance while maintaining a compliant impedance profile, in comparison to the variable impedance control (VIC) approach which was previously used in HRC literature for dealing with vibrations [6, 7]. We successfully applied the BMFLC algorithm [8] in our high-dof robotic arm for learning and suppressing the vibration online. In this work, we extend both our theoretical approach and experiments.