Catastrophic forgetting is a pervasive issue for pre-trained language models (PLMs) during continual learning, where models lose previously acquired knowledge when sequentially trained on a series of tasks. The model's ability to retain old tasks is referred to as stability, while its adaptability to new tasks is called plasticity. Therefore, the key to solving this problem is to find a trade-off between the plasticity and stability of the model. To address this issue, in this paper, we propose a novel method to achieve a balance between model stability and plasticity, thereby mitigating catastrophic forgetting. More specifically, our proposed approach leverages parameter isolation and a subsequent combination strategy. Initially, in the training stage, the model adapts to each downstream task via a parameter isolation method to prevent potential interference among different tasks. We then combine all trained parameters, which contain acquired knowledge, using the task arithmetic method and finally apply them to the backbone model. Empirical evaluations on continual language learning benchmarks substantiate the effectiveness of our approach, revealing a marked enhancement over existing state-of-the-art approaches.

Shanghai has officially unveiled its first heterogeneous humanoid robot training center, marking a significant accomplishment in China's robotics development. The Humanoid Robot Kylin Training Ground represents an important step in the country's technological advancement, showcasing China's commitment to becoming a global leader in robotics and artificial intelligence. The National and Local Co-Built Humanoid Robotics Innovation Center has launched a groundbreaking training facility that is revolutionizing the field of robotics. This cutting-edge complex, spanning over 53,800 square feet, is currently capable of training more than 100 humanoid robots at once. GET SECURITY ALERTS & EXPERT TECH TIPS – SIGN UP FOR KURT'S THE CYBERGUY REPORT NOW These advanced robots have showcased exceptional proficiency, with an average success rate exceeding 90% in various tasks.

Since it first appeared with the launch of match.com Around 10% of heterosexual people and 24% of LGBT people have met their long-term partner online, according to Pew Research Center. But evidence suggests that young people are switching off dating apps, with the UK's top 10 seeing a fall of nearly 16%, according to a report published by Ofcom in November 2024. Tinder lost 594,000 users, while Hinge dropped by 131,000, Bumble by 368,000 and Grindr by 11,000, the report said (a Grindr spokesperson said they were "not familiar with this study's source data" and that their UK users "continue to rise year over year"). According to a 2023 Axios study of US college students and other Gen Zers, 79% said they were forgoing regular dating app usage.

We introduce Agent Process Reward Models (AgentPRM), a simple and scalable framework for training LLM agents to continually improve through interactions. AgentPRM follows a lightweight actor-critic paradigm, using Monte Carlo rollouts to compute reward targets and optimize policies. It requires minimal modifications to existing RLHF pipelines, making it easy to integrate at scale. Beyond AgentPRM, we propose InversePRM, which learns process rewards directly from demonstrations without explicit outcome supervision. We also explore key challenges and opportunities, including exploration, process reward shaping, and model-predictive reasoning. We evaluate on ALFWorld benchmark, show that small 3B models trained with AgentPRM and InversePRM outperform strong GPT-4o baselines, and analyze test-time scaling, reward hacking, and more.

In this book, written in Portuguese, we discuss what ill-posed problems are and how the regularization method is used to solve them. In the form of questions and answers, we reflect on the origins and future of regularization, relating the similarities and differences of its meaning in different areas, including inverse problems, statistics, machine learning, and deep learning.

In this research, we explored the efficacy of various warning label designs for AI-generated content on social media platforms e.g., deepfakes. We devised and assessed ten distinct label design samples that varied across the dimensions of sentiment, color/iconography, positioning, and level of detail. Our experimental study involved 911 participants randomly assigned to these ten label designs and a control group evaluating social media content. We explored their perceptions relating to 1. Belief in the content being AI-generated, 2. Trust in the labels and 3. Social Media engagement perceptions of the content. The results demonstrate that the presence of labels had a significant effect on the users belief that the content is AI generated, deepfake, or edited by AI. However their trust in the label significantly varied based on the label design. Notably, having labels did not significantly change their engagement behaviors, such as like, comment, and sharing. However, there were significant differences in engagement based on content type: political and entertainment. This investigation contributes to the field of human computer interaction by defining a design space for label implementation and providing empirical support for the strategic use of labels to mitigate the risks associated with synthetically generated media.

We study the potential of large language models (LLMs) as proxies for humans to simplify preference elicitation (PE) in combinatorial assignment. While traditional PE methods rely on iterative queries to capture preferences, LLMs offer a one-shot alternative with reduced human effort. We propose a framework for LLM proxies that can work in tandem with SOTA ML-powered preference elicitation schemes. Our framework handles the novel challenges introduced by LLMs, such as response variability and increased computational costs. We experimentally evaluate the efficiency of LLM proxies against human queries in the well-studied course allocation domain, and we investigate the model capabilities required for success. We find that our approach improves allocative efficiency by up to 20%, and these results are robust across different LLMs and to differences in quality and accuracy of reporting.

Memory is crucial for enabling agents to tackle complex tasks with temporal and spatial dependencies. While many reinforcement learning (RL) algorithms incorporate memory, the field lacks a universal benchmark to assess an agent's memory capabilities across diverse scenarios. This gap is particularly evident in tabletop robotic manipulation, where memory is essential for solving tasks with partial observability and ensuring robust performance, yet no standardized benchmarks exist. To address this, we introduce MIKASA (Memory-Intensive Skills Assessment Suite for Agents), a comprehensive benchmark for memory RL, with three key contributions: (1) we propose a comprehensive classification framework for memory-intensive RL tasks, (2) we collect MIKASA-Base - a unified benchmark that enables systematic evaluation of memory-enhanced agents across diverse scenarios, and (3) we develop MIKASA-Robo - a novel benchmark of 32 carefully designed memory-intensive tasks that assess memory capabilities in tabletop robotic manipulation. Our contributions establish a unified framework for advancing memory RL research, driving the development of more reliable systems for real-world applications. The code is available at https://sites.google.com/view/memorybenchrobots/.

BabyLM aims to dissolve the boundaries between cognitive modeling and language modeling. We call for both workshop papers and for researchers to join the 3rd BabyLM competition. As in previous years, we call for participants in the data-efficient pretraining challenge in the general track. This year, we also offer a new track: INTERACTION. This new track encourages interactive behavior, learning from a teacher, and adapting the teaching material to the student. We also call for papers outside the competition in any relevant areas. These include training efficiency, cognitively plausible research, weak model evaluation, and more.

Recent advancements in this field have including the development of consistent surrogate losses for introduced features enabling flexibility to unseen training these systems (Mozannar & Sontag, 2021; Verma experts at test-time, but we find these approaches & Nalisnick, 2022), and extensions that allow for deferral have significant limitations. To address these, we to multiple experts (Verma et al., 2023). Recent work by introduce EA-L2D: Expert-Agnostic Learning to Tailor et al. (2024) proposed a meta-learning solution for Defer, a novel L2D framework that leverages a L2D systems that can adapt to experts not seen during the Bayesian approach to model expert behaviour in training regime through meta-learning representations of an expert-agnostic manner, facilitating optimal expert behaviours, enabling the system to quickly adapt to deferral decisions. EA-L2D offers several critical new experts using a small set of their example predictions, improvements over prior methods, including denoted context predictions. However, this approach exhibits the ability to incorporate prior knowledge about a key weakness in limited generalisation to experts experts, a reduced reliance on expert-annotated with expertise unseen during training. Additionally, their data, and robust performance when deferring to solution poses problems seen more widely in L2D literature, experts with expertise not seen during training.