Personalization in large language models (LLMs) seeks to tailor models to individual user or user group preferences. Prompt-based methods augment queries with user preference information, whereas training-based methods directly encode preferences into model parameters for more effective personalization. Despite achieving some success in personalizing LLMs, prior methods often fail to directly optimize task-specific metrics and lack explicit privacy-preservation mechanisms. To address these limitations, we propose Privacy-Preserving Model Merging via Evolutionary Algorithms (PriME), a novel approach to personalization that employs gradient-free methods to directly optimize task-specific metrics while preserving user privacy. By incorporating privacy preservation into optimization, PriME produces a personalized module that effectively captures the target user's preferences while minimizing the privacy risks for the users sharing their private information. Experiments on the LaMP benchmark show that PriME outperforms both prompt-based and training-based methods, achieving up to a 45% performance improvement over the prior art. Further analysis shows that PriME achieves a significantly better privacy-utility trade-off, highlighting the potential of evolutionary approaches for privacy-preserving LLM personalization.

Large Language Models (LLMs) have achieved remarkable success, but their English-centric training data limits performance in non-English languages, highlighting the need for enhancements in their multilingual capabilities. While some work on multilingual prompting methods handles non-English queries by utilizing English translations or restructuring them to more closely align with LLM reasoning patterns, these works often overlook the importance of cultural context, limiting their effectiveness. To address this limitation, we propose EMCEI, a simple yet effective approach that improves LLMs' multilingual capabilities by incorporating cultural context for more accurate and appropriate responses. Specifically, EMCEI follows a two-step process that first extracts relevant cultural context from the LLM's parametric knowledge via prompting. Then, EMCEI employs an LLM-as-Judge mechanism to select the most appropriate response by balancing cultural relevance and reasoning ability. Experiments on diverse multilingual benchmarks show that EMCEI outperforms existing baselines, demonstrating its effectiveness in handling multilingual queries with LLMs.

--Dynamic and contact-rich object manipulation, such as striking, snatching, or hammering, remains challenging for robotic systems due to hardware limitations. Most existing robots are constrained by high-inertia design, limited compliance, and reliance on expensive torque sensors. T o address this, we introduce ARMADA (Affordable Robot for Manipulation and Dynamic Actions), a 6 degrees-of-freedom bimanual robot designed for dynamic manipulation research. ARMADA combines low-inertia, back-drivable actuators with a lightweight design, using readily available components and 3D-printed links for ease of assembly in research labs. The entire system, including both arms, is built for just $6,100. Each arm achieves speeds up to 6.16m/s, almost twice that of most collaborative robots, with a comparable payload of 2.5kg. We demonstrate ARMADA can perform dynamic manipulation like snatching, hammering, and bimanual throwing in real-world environments. We also showcase its effectiveness in reinforcement learning (RL) by training a non-prehensile manipulation policy in simulation and transferring it zero-shot to the real world, as well as human motion shadowing for dynamic bimanual object throwing. ARMADA is fully open-sourced with detailed assembly instructions, CAD models, URDFs, simulation, and learning codes. We highly recommend viewing the supplementary video at https://sites.google.com/view/im2-humanoid-arm. I NTRODUCTION Humans use a rich set of action repertoire to manipulate objects: we not only pick-and-place objects but also toss laundry, slide a box, snatch a pen, hammer a nail, otherwise arXiv:2502.16908v1 In contrast, most manipulators today are limited to picking, where a robot simply grasps an object to resist the frictional force. While kinematic pick-and-place is sufficient for static, controlled tasks, dynamic manipulation is necessary for building an effective general-purpose robot.

Self-awareness, i.e., the ability to assess and correct one's own generation, is a fundamental aspect of human intelligence, making its replication in large language models (LLMs) an important yet challenging task. Previous works tackle this by employing extensive reinforcement learning or rather relying on large external verifiers. In this work, we propose Refine via Intrinsic Self-Verification (ReVISE), an efficient and effective framework that enables LLMs to self-correct their outputs through self-verification. The core idea of ReVISE is to enable LLMs to verify their reasoning processes and continually rethink reasoning trajectories based on its verification. We introduce a structured curriculum based upon online preference learning to implement this efficiently. Specifically, as ReVISE involves two challenging tasks (i.e., self-verification and reasoning correction), we tackle each task sequentially using curriculum learning, collecting both failed and successful reasoning paths to construct preference pairs for efficient training. During inference, our approach enjoys natural test-time scaling by integrating self-verification and correction capabilities, further enhanced by our proposed confidence-aware decoding mechanism. Our experiments on various reasoning tasks demonstrate that ReVISE achieves efficient self-correction and significantly improves reasoning performance.

The adaptation of large language models (LLMs) to chemistry has shown promising performance in molecular understanding tasks, such as generating a text description from a molecule. However, proper reasoning based on molecular structural information remains a significant challenge, e.g., even advanced LLMs such as GPT-4o struggle to identify functional groups which are crucial for inferring the molecular property of interest. To address this limitation, we propose StructCoT, a structure-aware chain-of-thought (CoT) that enhances LLMs' understanding of molecular structures by explicitly injecting the key structural features of molecules. Moreover, we introduce two fine-tuning frameworks for adapting the existing LLMs to use our StructCoT. Our experiments demonstrate that incorporating StructCoT with our fine-tuning frameworks leads to consistent improvements in both molecular understanding tasks.

Recent advancements in large language models (LLMs) have demonstrated impressive performance in generating molecular structures as drug candidates, which offers significant potential to accelerate drug discovery. However, the current LLMs overlook a critical requirement for drug discovery: proposing a diverse set of molecules. This diversity is essential for improving the chances of finding a viable drug, as it provides alternative molecules that may succeed where others fail in wet-lab or clinical validations. Despite such a need for diversity, the LLMs often output structurally similar molecules from a given prompt. While decoding schemes like beam search may enhance textual diversity, this often does not align with molecular structural diversity. In response, we propose a new method for fine-tuning molecular generative LLMs to autoregressively generate a set of structurally diverse molecules, where each molecule is generated by conditioning on the previously generated molecules. Our approach consists of two stages: (1) supervised fine-tuning to adapt LLMs to autoregressively generate molecules in a sequence and (2) reinforcement learning to maximize structural diversity within the generated molecules. Our experiments show that (1) our fine-tuning approach enables the LLMs to better discover diverse molecules compared to existing decoding schemes and (2) our fine-tuned model outperforms other representative LLMs in generating diverse molecules, including the ones fine-tuned on chemical domains.

As the diversity of users increases, the capability of providing personalized responses by large language models (LLMs) has become increasingly important. Existing approaches have only limited successes in LLM personalization, due to the absence of personalized learning or the reliance on shared personal data. This paper proposes a new approach for a few-shot personalization of LLMs with their mis-aligned responses (Fermi). Our key idea is to learn a set of personalized prompts for each user by progressively improving the prompts using LLMs, based on user profile (e.g., demographic information) and a few examples of previous opinions. During an iterative process of prompt improvement, we incorporate the contexts of mis-aligned responses by LLMs, which are especially crucial for the effective personalization of LLMs. In addition, we develop an effective inference method to further leverage the context of the test query and the personalized prompts. Our experimental results demonstrate that Fermi significantly improves performance across various benchmarks, compared to the best-performing baselines.

An open challenge in recent machine learning is about how to improve the reasoning capability of large language models (LLMs) in a black-box setting, i.e., without access to detailed information such as output token probabilities. Existing approaches either rely on accessibility (which is often unrealistic) or involve significantly increased train- and inference-time costs. This paper addresses those limitations or shortcomings by proposing a novel approach, namely CoBB (Correct for improving QA reasoning of Black-Box LLMs). It uses a trained adaptation model to perform a seq2seq mapping from the often-imperfect reasonings of the original black-box LLM to the correct or improved reasonings. Specifically, the adaptation model is initialized with a relatively small open-source LLM and adapted over a collection of sub-sampled training pairs. To select the representative pairs of correct and incorrect reasonings, we formulated the dataset construction as an optimization problem that minimizes the statistical divergence between the sampled subset and the entire collection, and solved it via a genetic algorithm. We then train the adaptation model over the sampled pairs by contrasting the likelihoods of correct and incorrect reasonings. Our experimental results demonstrate that CoBB significantly improves reasoning accuracy across various QA benchmarks, compared to the best-performing adaptation baselines.

Learning effective representations from raw data is crucial for the success of deep learning methods. However, in the tabular domain, practitioners often prefer augmenting raw column features over using learned representations, as conventional tree-based algorithms frequently outperform competing approaches. As a result, feature engineering methods that automatically generate candidate features have been widely used. While these approaches are often effective, there remains ambiguity in defining the space over which to search for candidate features. Moreover, they often rely solely on validation scores to select good features, neglecting valuable feedback from past experiments that could inform the planning of future experiments. To address the shortcomings, we propose a new tabular learning framework based on large language models (LLMs), coined Optimizing Column feature generator with decision Tree reasoning (OCTree). Our key idea is to leverage LLMs' reasoning capabilities to find good feature generation rules without manually specifying the search space and provide language-based reasoning information highlighting past experiments as feedback for iterative rule improvements. Here, we choose a decision tree as reasoning as it can be interpreted in natural language, effectively conveying knowledge of past experiments (i.e., the prediction models trained with the generated features) to the LLM. Our empirical results demonstrate that this simple framework consistently enhances the performance of various prediction models across diverse tabular benchmarks, outperforming competing automatic feature engineering methods.

Aligning large language models (LLMs) with human preferences becomes a key component to obtaining state-of-the-art performance, but it yields a huge cost to construct a large human-annotated preference dataset. To tackle this problem, we propose a new framework that boosts the alignment of LLMs through Self-generated Preference data (Selfie) using only a very small amount of human-annotated preference data. Our key idea is leveraging the human prior knowledge within the small (seed) data and progressively improving the alignment of LLM, by iteratively generating the responses and learning from them with the self-annotated preference data. To be specific, we propose to derive the preference label from the logits of LLM to explicitly extract the model's inherent preference. Compared to the previous approaches using external reward models or implicit in-context learning, we observe that the proposed approach is significantly more effective. In addition, we introduce a noise-aware preference learning algorithm to mitigate the risk of low quality within generated preference data. Our experimental results demonstrate that the proposed framework significantly boosts the alignment of LLMs. For example, we achieve superior alignment performance on AlpacaEval 2.0 with only 3.3\% of the ground-truth preference labels in the Ultrafeedback data compared to the cases using the entire data or state-of-the-art baselines.