Archives AI News

arXiv:2505.08299v2 Announce Type: replace Abstract: State-space models (SSMs), particularly the Mamba architecture, have emerged as powerful alternatives to Transformers for sequence modeling, offering linear-time complexity and competitive performance across diverse tasks. However, their large parameter counts pose significant challenges for deployment in resource-constrained environments. We propose a novel unstructured pruning framework tailored for Mamba models that achieves up to 70% parameter reduction while retaining over 95% of the original performance. Our approach integrates three key innovations: (1) a gradient-aware magnitude pruning technique that combines weight magnitude and gradient information to identify less critical parameters, (2) an iterative pruning schedule that gradually increases sparsity to maintain model stability, and (3) a global pruning strategy that optimizes parameter allocation across the entire model. Through extensive experiments on WikiText-103, Long Range Arena, and ETT time-series benchmarks, we demonstrate significant efficiency gains with minimal performance degradation. Our analysis of pruning effects on Mamba's components reveals critical insights into the architecture's redundancy and robustness, enabling practical deployment in resource-constrained settings while broadening Mamba's applicability.

arXiv:2509.05489v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards has significantly advanced reasoning in large language models (LLMs), but such signals remain coarse, offering only binary correctness feedback. This limitation often results in inefficiencies, including overly verbose reasoning and high computational cost, while existing solutions often compromise accuracy. To address this, we introduce self-aligned reward (SAR), a self-guided signal that complements verifiable rewards to encourage both reasoning accuracy and efficiency. SAR is defined as the relative perplexity difference between an answer conditioned on the query and the standalone answer, thereby favoring responses that are concise and query-specific. Quantitative analysis reveals that SAR reliably distinguishes answer quality: concise, correct answers score higher than redundant ones, and partially correct answers score higher than entirely incorrect ones. Evaluation on 4 models across 7 benchmarks shows that integrating SAR with prevalent RL algorithms like PPO and GRPO improves accuracy by 4%, while reducing inference cost by 30%. Further analysis demonstrates that SAR achieves a Pareto-optimal trade-off between correctness and efficiency compared to reward signals based on length or self-confidence. We also show that SAR shortens responses while preserving advanced reasoning behaviors, demonstrating its ability to suppress unnecessary elaboration without losing critical reasoning. These results highlight the promise of self-aligned reward as a fine-grained complement to verifiable rewards, paving the way for more efficient and effective LLM training.

arXiv:2507.07032v2 Announce Type: replace Abstract: Protein structure prediction often hinges on multiple sequence alignments (MSAs), which underperform on low-homology and orphan proteins. We introduce PLAME, a lightweight MSA design framework that leverages evolutionary embeddings from pretrained protein language models to generate MSAs that better support downstream folding. PLAME couples these embeddings with a conservation-diversity loss that balances agreement on conserved positions with coverage of plausible sequence variation. Beyond generation, we develop (i) an MSA selection strategy to filter high-quality candidates and (ii) a sequence-quality metric that is complementary to depth-based measures and predictive of folding gains. On AlphaFold2 low-homology/orphan benchmarks, PLAME delivers state-of-the-art improvements in structure accuracy (e.g., lDDT/TM-score), with consistent gains when paired with AlphaFold3. Ablations isolate the benefits of the selection strategy, and case studies elucidate how MSA characteristics shape AlphaFold confidence and error modes. Finally, we show PLAME functions as a lightweight adapter, enabling ESMFold to approach AlphaFold2-level accuracy while retaining ESMFold-like inference speed. PLAME thus provides a practical path to high-quality folding for proteins lacking strong evolutionary neighbors.

arXiv:2509.05542v1 Announce Type: new Abstract: Training multimodal process reward models (PRMs) is challenged by distribution shifts and noisy data. We introduce DreamPRM-1.5, an instance-reweighted framework that adaptively adjusts the importance of each training example via bi-level optimization. We design two complementary strategies: Instance Table, effective for smaller datasets, and Instance Net, scalable to larger ones. Integrated into test-time scaling, DreamPRM-1.5 achieves 84.6 accuracy on the MMMU benchmark, surpassing GPT-5.

arXiv:2402.12391v3 Announce Type: replace-cross Abstract: Machine learning has emerged as a powerful tool for scientific discovery, enabling researchers to extract meaningful insights from complex datasets. For instance, it has facilitated the identification of disease-predictive genes from gene expression data, significantly advancing healthcare. However, the traditional process for analyzing such datasets demands substantial human effort and expertise for the data selection, processing, and analysis. To address this challenge, we introduce a novel framework, a Team of AI-made Scientists (TAIS), designed to streamline the scientific discovery pipeline. TAIS comprises simulated roles, including a project manager, data engineer, and domain expert, each represented by a Large Language Model (LLM). These roles collaborate to replicate the tasks typically performed by data scientists, with a specific focus on identifying disease-predictive genes. Furthermore, we have curated a benchmark dataset to assess TAIS's effectiveness in gene identification, demonstrating our system's potential to significantly enhance the efficiency and scope of scientific exploration. Our findings represent a solid step towards automating scientific discovery through large language models.

arXiv:2509.05545v1 Announce Type: new Abstract: Solving long-horizon goal-conditioned tasks remains a significant challenge in reinforcement learning (RL). Hierarchical reinforcement learning (HRL) addresses this by decomposing tasks into more manageable sub-tasks, but the automatic discovery of the hierarchy and the joint training of multi-level policies often suffer from instability and can lack theoretical guarantees. In this paper, we introduce Reinforcement Learning with Anticipation (RLA), a principled and potentially scalable framework designed to address these limitations. The RLA agent learns two synergistic models: a low-level, goal-conditioned policy that learns to reach specified subgoals, and a high-level anticipation model that functions as a planner, proposing intermediate subgoals on the optimal path to a final goal. The key feature of RLA is the training of the anticipation model, which is guided by a principle of value geometric consistency, regularized to prevent degenerate solutions. We present proofs that RLA approaches the globally optimal policy under various conditions, establishing a principled and convergent method for hierarchical planning and execution in long-horizon goal-conditioned tasks.

arXiv:2502.03365v2 Announce Type: replace-cross Abstract: Software vulnerabilities are commonly detected via static analysis, penetration testing, and fuzzing. They can also be found by running unit tests - so-called vulnerability-witnessing tests - that stimulate the security-sensitive behavior with crafted inputs. Developing such tests is difficult and time-consuming; thus, automated data-driven approaches could help developers intercept vulnerabilities earlier. However, training and validating such approaches require a lot of data, which is currently scarce. This paper introduces VUTECO, a deep learning-based approach for collecting instances of vulnerability-witnessing tests from Java repositories. VUTECO carries out two tasks: (1) the "Finding" task to determine whether a test case is security-related, and (2) the "Matching" task to relate a test case to the exact vulnerability it is witnessing. VUTECO successfully addresses the Finding task, achieving perfect precision and 0.83 F0.5 score on validated test cases in VUL4J and returning 102 out of 145 (70%) correct security-related test cases from 244 open-source Java projects. Despite showing sufficiently good performance for the Matching task - i.e., 0.86 precision and 0.68 F0.5 score - VUTECO failed to retrieve any valid match in the wild. Nevertheless, we observed that in almost all of the matches, the test case was still security-related despite being matched to the wrong vulnerability. In the end, VUTECO can help find vulnerability-witnessing tests, though the matching with the right vulnerability is yet to be solved; the findings obtained lay the stepping stone for future research on the matter.

arXiv:2509.05584v1 Announce Type: new Abstract: Foundation models face growing compute and memory bottlenecks, hindering deployment on resource-limited platforms. While compression techniques such as pruning and quantization are widely used, most rely on uniform heuristics that ignore architectural and runtime heterogeneity. Profiling tools expose per-layer latency, memory, and compute cost, yet are rarely integrated into automated pipelines. We propose ProfilingAgent, a profiling-guided, agentic approach that uses large language models (LLMs) to automate compression via structured pruning and post-training dynamic quantization. Our modular multi-agent system reasons over static metrics (MACs, parameter counts) and dynamic signals (latency, memory) to design architecture-specific strategies. Unlike heuristic baselines, ProfilingAgent tailors layer-wise decisions to bottlenecks. Experiments on ImageNet-1K, CIFAR-10, and CIFAR-100 with ResNet-101, ViT-B/16, Swin-B, and DeiT-B/16 show pruning maintains competitive or improved accuracy (about 1% drop on ImageNet-1K, +2% gains for ViT-B/16 on smaller datasets), while quantization achieves up to 74% memory savings with <0.5% accuracy loss. Our quantization also yields consistent inference speedups of up to 1.74 times faster. Comparative studies with GPT-4o and GPT-4-Turbo highlight the importance of LLM reasoning quality for iterative pruning. These results establish agentic systems as scalable solutions for profiling-guided model optimization.

arXiv:2509.05615v1 Announce Type: new Abstract: Medical multimodal representation learning aims to integrate heterogeneous clinical data into unified patient representations to support predictive modeling, which remains an essential yet challenging task in the medical data mining community. However, real-world medical datasets often suffer from missing modalities due to cost, protocol, or patient-specific constraints. Existing methods primarily address this issue by learning from the available observations in either the raw data space or feature space, but typically neglect the underlying bias introduced by the data acquisition process itself. In this work, we identify two types of biases that hinder model generalization: missingness bias, which results from non-random patterns in modality availability, and distribution bias, which arises from latent confounders that influence both observed features and outcomes. To address these challenges, we perform a structural causal analysis of the data-generating process and propose a unified framework that is compatible with existing direct prediction-based multimodal learning methods. Our method consists of two key components: (1) a missingness deconfounding module that approximates causal intervention based on backdoor adjustment and (2) a dual-branch neural network that explicitly disentangles causal features from spurious correlations. We evaluated our method in real-world public and in-hospital datasets, demonstrating its effectiveness and causal insights.

arXiv:2509.06796v1 Announce Type: cross Abstract: A Membership Inference Attack (MIA) assesses how much a target machine learning model reveals about its training data by determining whether specific query instances were part of the training set. State-of-the-art MIAs rely on training hundreds of shadow models that are independent of the target model, leading to significant computational overhead. In this paper, we introduce Imitative Membership Inference Attack (IMIA), which employs a novel imitative training technique to strategically construct a small number of target-informed imitative models that closely replicate the target model's behavior for inference. Extensive experimental results demonstrate that IMIA substantially outperforms existing MIAs in various attack settings while only requiring less than 5% of the computational cost of state-of-the-art approaches.