🤖 AI资讯日报

AI领域的下一个比尔·盖茨或爱因斯坦

探讨AI进步速度是否呈指数级增长

云计算GPU成本降低50%，相比AWS节省开发者50%

探讨当前AI算法的痛点

NLP、AI、ML、机器人是短暂趋势还是更多？

谷歌提供Common Lisp与机器学习实习机会

初探人工智能及其预期

AI领域的疯狂指数讨论

关于研究生院的思考（CS博士）

讨论对纽约市地方法律144的关注

初创公司通过书籍销售筹集资金

生物信息学家的讨论

综述自进化代理，迈向人工超级智能的道路。

GenoMAS：通过代码驱动的基因表达分析进行科学发现的多代理框架。

Deep Neural Networks (DNNs) deliver impressive performance but their black-box nature limits deployment in high-stakes domains requiring transparency. We introduce Compositional Function Networks (CFNs), a novel framework that builds inherently interpretable models by composing elementary mathematical functions with clear semantics. Unlike existing interpretable approaches that are limited to simple additive structures, CFNs support diverse compositional patterns -- sequential, parallel, and conditional -- enabling complex feature interactions while maintaining transparency. A key innovation is that CFNs are fully differentiable, allowing efficient training through standard gradient descent. We demonstrate CFNs' versatility across multiple domains, from symbolic regression to image classification with deep hierarchical networks. Our empirical evaluation shows CFNs achieve competitive performance against black-box models (96.24% accuracy on CIFAR-10) while outperforming state-of-the-art interpretable models like Explainable Boosting Machines. By combining the hierarchical expressiveness and efficient training of deep learning with the intrinsic interpretability of well-defined mathematical functions, CFNs offer a powerful framework for applications where both performance and accountability are paramount.

While frontier large language models (LLMs) continue to push capability boundaries, their deployment remains confined to GPU-powered cloud infrastructure. We challenge this paradigm with SmallThinker, a family of LLMs natively designed - not adapted - for the unique constraints of local devices: weak computational power, limited memory, and slow storage. Unlike traditional approaches that mainly compress existing models built for clouds, we architect SmallThinker from the ground up to thrive within these limitations. Our innovation lies in a deployment-aware architecture that transforms constraints into design principles. First, We introduce a two-level sparse structure combining fine-grained Mixture-of-Experts (MoE) with sparse feed-forward networks, drastically reducing computational demands without sacrificing model capacity. Second, to conquer the I/O bottleneck of slow storage, we design a pre-attention router that enables our co-designed inference engine to prefetch expert parameters from storage while computing attention, effectively hiding storage latency that would otherwise cripple on-device inference. Third, for memory efficiency, we utilize NoPE-RoPE hybrid sparse attention mechanism to slash KV cache requirements. We release SmallThinker-4B-A0.6B and SmallThinker-21B-A3B, which achieve state-of-the-art performance scores and even outperform larger LLMs. Remarkably, our co-designed system mostly eliminates the need for expensive GPU hardware: with Q4_0 quantization, both models exceed 20 tokens/s on ordinary consumer CPUs, while consuming only 1GB and 8GB of memory respectively. SmallThinker is publicly available at hf.co/PowerInfer/SmallThinker-4BA0.6B-Instruct and hf.co/PowerInfer/SmallThinker-21BA3B-Instruct.

MIRAGE-Bench：LLM代理的幻觉及其发现位置。

In real-world machine learning deployments, models must be continually updated, composed, and when required, selectively undone. However, existing approaches to model merging and continual learning often suffer from task interference, catastrophic forgetting, or lack of reversibility. We propose Modular Delta Merging with Orthogonal Constraints (MDM-OC), a novel framework that enables scalable, interference-free, and reversible composition of fine-tuned models. Each task-specific model is encoded as a delta from a shared base and projected into an orthogonal subspace to eliminate conflict. These projected deltas are then merged via gradient-based optimization to form a unified model that retains performance across tasks. Our approach supports continual integration of new models, structured unmerging for compliance such as GDPR requirements, and model stability via elastic weight consolidation and synthetic replay. Extensive experiments on vision and natural language processing benchmarks demonstrate that MDM-OC outperforms prior baselines in accuracy, backward transfer, and unmerge fidelity, while remaining memory-efficient and computationally tractable. This framework offers a principled solution for modular and compliant AI system design.

Large visual-language models (LVLMs) integrate aligned large language models (LLMs) with visual modules to process multimodal inputs. However, the safety mechanisms developed for text-based LLMs do not naturally extend to visual modalities, leaving LVLMs vulnerable to harmful image inputs. To address this cross-modal safety gap, we introduce security tensors - trainable input vectors applied during inference through either the textual or visual modality. These tensors transfer textual safety alignment to visual processing without modifying the model's parameters. They are optimized using a curated dataset containing (i) malicious image-text pairs requiring rejection, (ii) contrastive benign pairs with text structurally similar to malicious queries, with the purpose of being contrastive examples to guide visual reliance, and (iii) general benign samples preserving model functionality. Experimental results demonstrate that both textual and visual security tensors significantly enhance LVLMs' ability to reject diverse harmful visual inputs while maintaining near-identical performance on benign tasks. Further internal analysis towards hidden-layer representations reveals that security tensors successfully activate the language module's textual "safety layers" in visual inputs, thereby effectively extending text-based safety to the visual modality.

基于ASP的交互式配置的智能扩展技术。

This study investigates the mechanisms and factors influencing memorization in fine-tuned large language models (LLMs), with a focus on the medical domain due to its privacy-sensitive nature. We examine how different aspects of the fine-tuning process affect a model's propensity to memorize training data, using the PHEE dataset of pharmacovigilance events. Our research employs two main approaches: a membership inference attack to detect memorized data, and a generation task with prompted prefixes to assess verbatim reproduction. We analyze the impact of adapting different weight matrices in the transformer architecture, the relationship between perplexity and memorization, and the effect of increasing the rank in low-rank adaptation (LoRA) fine-tuning. Key findings include: (1) Value and Output matrices contribute more significantly to memorization compared to Query and Key matrices; (2) Lower perplexity in the fine-tuned model correlates with increased memorization; (3) Higher LoRA ranks lead to increased memorization, but with diminishing returns at higher ranks. These results provide insights into the trade-offs between model performance and privacy risks in fine-tuned LLMs. Our findings have implications for developing more effective and responsible strategies for adapting large language models while managing data privacy concerns.

Existing methods for estimating personalized treatment effects typically rely on structured covariates, limiting their applicability to unstructured data. Yet, leveraging unstructured data for causal inference has considerable application potential, for instance in healthcare, where clinical notes or medical images are abundant. To this end, we first introduce an approximate 'plug-in' method trained directly on the neural representations of unstructured data. However, when these fail to capture all confounding information, the method may be subject to confounding bias. We therefore introduce two theoretically grounded estimators that leverage structured measurements of the confounders during training, but allow estimating personalized treatment effects purely from unstructured inputs, while avoiding confounding bias. When these structured measurements are only available for a non-representative subset of the data, these estimators may suffer from sampling bias. To address this, we further introduce a regression-based correction that accounts for the non-uniform sampling, assuming the sampling mechanism is known or can be well-estimated. Our experiments on two benchmark datasets show that the plug-in method, directly trainable on large unstructured datasets, achieves strong empirical performance across all settings, despite its simplicity.

Domain shift poses a fundamental challenge in time series analysis, where models trained on source domain often fail dramatically when applied in target domain with different yet similar distributions. While current unsupervised domain adaptation (UDA) methods attempt to align cross-domain feature distributions, they typically treat features as indivisible entities, ignoring their intrinsic compositions that governs domain adaptation. We introduce DARSD, a novel UDA framework with theoretical explainability that explicitly realizes UDA tasks from the perspective of representation space decomposition. Our core insight is that effective domain adaptation requires not just alignment, but principled disentanglement of transferable knowledge from mixed representations. DARSD consists three synergistic components: (I) An adversarial learnable common invariant basis that projects original features into a domain-invariant subspace while preserving semantic content; (II) A prototypical pseudo-labeling mechanism that dynamically separates target features based on confidence, hindering error accumulation; (III) A hybrid contrastive optimization strategy that simultaneously enforces feature clustering and consistency while mitigating emerging distribution gaps. Comprehensive experiments conducted on four benchmark datasets (WISDM, HAR, HHAR, and MFD) demonstrate DARSD's superiority against 12 UDA algorithms, achieving optimal performance in 35 out of 53 cross-domain scenarios.

Recent advances in diffusion-based video generation have enabled photo-realistic short clips, but current methods still struggle to achieve multi-modal consistency when jointly generating whole-body motion and natural speech. Current approaches lack comprehensive evaluation frameworks that assess both visual and audio quality, and there are insufficient benchmarks for region-specific performance analysis. To address these gaps, we introduce the Joint Whole-Body Talking Avatar and Speech Generation Version I(JWB-DH-V1), comprising a large-scale multi-modal dataset with 10,000 unique identities across 2 million video samples, and an evaluation protocol for assessing joint audio-video generation of whole-body animatable avatars. Our evaluation of SOTA models reveals consistent performance disparities between face/hand-centric and whole-body performance, which incidates essential areas for future research. The dataset and evaluation tools are publicly available at https://github.com/deepreasonings/WholeBodyBenchmark.