The Mamba model has gained significant attention for its computational advantages over Transformer-based models, while achieving comparable performance across a wide range of language tasks. Like Transformers, Mamba exhibits in-context learning (ICL) capabilities, i.e., making predictions for new tasks based on a prompt containing input-label pairs and a query, without requiring fine-tuning. Despite its empirical success, the theoretical understanding of Mamba remains limited, largely due to the nonlinearity introduced by its gating mechanism. To the best of our knowledge, this paper presents the first theoretical analysis of the training dynamics of a one-layer Mamba model, which consists of a linear attention component followed by a nonlinear gating layer, and its ICL generalization on unseen binary classification tasks, even when the prompt includes additive outliers. Our analysis shows that Mamba leverages the linear attention layer to select informative context examples and uses the nonlinear gating layer to suppress the influence of outliers. By establishing and comparing to the analysis of linear Transformers under the same setting, we show that although Mamba may require more training iterations to converge, it maintains accurate predictions even when the proportion of outliers exceeds the threshold that a linear Transformer can tolerate. These theoretical findings are supported by empirical experiments.

Epilepsy is a chronic neurological condition characterized by recurrent seizures, with global prevalence estimated at 50 million people worldwide. While progress in high-throughput sequencing has allowed for broad-based transcriptomic profiling of brain tissues, the deciphering of these highly complex datasets remains one of the challenges. To address this issue, in this paper we propose a new analysis pipeline that integrates the power of deep learning strategies with GPU-acceleration computation for investigating Gene expression patterns in epilepsy. Specifically, our proposed approach employs GPT-2 XL, a transformer-based Large Language Model (LLM) with 1.5 billion parameters for genomic sequence analysis over the latest NVIDIA H100 Tensor Core GPUs based on Hopper architecture. Our proposed method enables efficient preprocessing of RNA sequence data, gene sequence encoding, and subsequent pattern identification. We conducted experiments on two epilepsy datasets including GEO accession GSE264537 and GSE275235. The obtained results reveal several significant transcriptomic modifications, including reduced hippocampal astrogliosis after ketogenic diet treatment as well as restored excitatory-inhibitory signaling equilibrium in zebrafish epilepsy model. Moreover, our results highlight the effectiveness of leveraging LLMs in combination with advanced hardware acceleration for transcriptomic characterization in neurological diseases.

Tensor networks (TNs) enable compact representations of large tensors through shared parameters. Their use in probabilistic modeling is particularly appealing, as probabilistic tensor networks (PTNs) allow for tractable computation of marginals. However, existing approaches for learning parameters of PTNs are either computationally demanding and not fully compatible with automatic differentiation frameworks, or numerically unstable. In this work, we propose a conceptually simple approach for learning PTNs efficiently, that is numerically stable. We show our method provides significant improvements in time and space complexity, achieving 10x reduction in latency for generative modeling on the MNIST dataset. Furthermore, our approach enables learning of distributions with 10x more variables than previous approaches when applied to a variety of density estimation benchmarks. Our code is publicly available at github.com/marawangamal/ptn.

We present the transformer cookbook: a collection of techniques for directly encoding algorithms into a transformer's parameters. This work addresses the steep learning curve of such endeavors, a problem exacerbated by a fragmented literature where key results are scattered across numerous papers. In particular, we synthesize this disparate body of findings into a curated set of recipes that demonstrate how to implement everything from basic arithmetic in feed-forward layers to complex data routing via self-attention. Our mise en place of formulations is for both newcomers seeking an accessible entry point and experts in need of a systematic reference. This unified presentation of transformer constructions provides a foundation for future work spanning theoretical research in computational complexity to empirical investigations in architecture design and interpretability.

AI question answering systems increasingly generate responses with attributions to sources. However, the task of verifying the actual content of these attributions is in most cases impractical. In this paper, we present attribution gradients as a solution. Attribution gradients provide integrated, incremental affordances for diving into an attributed passage. A user can decompose a sentence of an answer into its claims. For each claim, the user can view supporting and contradictory excerpts mined from sources. Those excerpts serve as clickable conduits into the source (in our application, scientific papers). When evidence itself contains more citations, the UI unpacks the evidence into excerpts from the cited sources. These features of attribution gradients facilitate concurrent interconnections among answer, claim, excerpt, and context. In a usability study, we observed greater engagement with sources and richer revision in a task where participants revised an attributed AI answer with attribution gradients and a baseline.

Protein structure tokenizers enable the creation of multimodal models of protein structure, sequence, and function. Current approaches to protein structure tok-enization rely on bespoke components that are invariant to spatial symmetries, but that are challenging to optimize and scale. Kanzi consists of a diffusion autoencoder trained with a flow matching loss. We show that this approach simplifies several aspects of protein structure tokenizers: frame-based representations can be replaced with global coordinates, complex losses are replaced with a single flow matching loss, and SE(3)-invariant attention operations can be replaced with standard attention. We find that these changes stabilize the training of parameter-efficient models that outperform existing to-kenizers on reconstruction metrics at a fraction of the model size and training cost. An autoregressive model trained with Kanzi outperforms similar generative models that operate over tokens, although it does not yet match the performance of state-of-the-art continuous diffusion models. Code is available here: https://github.com/rdilip/kanzi/. The promise of digital biology is to develop machine learning models that are capable of performing a wide range of tasks, from generating novel therapeutics to reasoning about cellular-level processes (Richardson & Richardson, 1989; Cui et al., 2025; Kuhlman & Bradley, 2019). Proteins, which are essential components of almost all biological processes, are a natural target for machine learning approaches to biological perception and generation. A recent exciting advance has been the development of multimodal deep learning models capable of reasoning over protein sequence, structure, and function (Wang et al., 2024; Zhang et al., 2024; Liu et al., 2024; Gaujac et al., 2024; Hayes et al., 2025). These models are enabled by structure tokenizers, which convert the continuous three-dimensional protein structures into a sequence of discrete tokens from a finite vocabulary using vector quantization (V an Den Oord et al., 2017). Training language models on these discrete token sequences unlocks the possibility of truly multimodal biological models that excel across representation and generative tasks. Work done while not at OpenAI. Work done prior to the author's employment at Amazon. These SE(3)-invariant modules, however, can be challenging to both optimize at scale and extend to more diverse biological molecules (e.g., proteins with post-translational modifications, RNA, DNA). Models that can accurately tokenize protein structures without explicitly encoding spatial symmetries may offer improved flexibility and scalability for modeling biology. Such models, however, do not currently exist. Bridging this gap and exploring the performance of non-invariant protein tokenizers is the primary focus of this work.

Adaptive chatbots that mimic a user's linguistic style can build rapport and engagement, yet unconstrained mimicry risks an agent that feels unstable or sycophantic. We present a computational evaluation framework that makes the core design tension explicit: balancing moment-to-moment linguistic synchrony against long-term persona stability. Using an 8-dimensional style vector and a closed-loop "base+delta" prompting architecture, we simulate and compare explicit adaptation policies - Uncapped, Cap, Exponential Moving Average (EMA), Dead-Band, and Hybrids - on a human-log dataset. Our analysis maps a clear Pareto frontier: bounded policies achieve substantial gains in stability at a modest cost to synchrony. For example, a Hybrid (EMA+Cap) raises stability from 0.542 to 0.878 (+62%) while reducing synchrony by only 17%. We confirm this trade-off through large-scale replications on three public corpora (DailyDialog, Persona-Chat, EmpatheticDialogues) and LLM-in-the-loop validation across two model families. Furthermore, we quantify "prompt legibility," showing that frontier policies reduce instruction churn and cut jarring register flips (major tone changes) from 0.254 to 0.092, yielding systems that are easier to reason about and maintain. Taken together, our framework provides a general evaluation harness for style adaptation; a systematic ablation that identifies Pareto-efficient policies; robust validation across diverse datasets and models; and novel legibility metrics linking policy choices to system maintainability.

Code search is an important information retrieval application. Benefits of better code search include faster new developer on-boarding, reduced software maintenance, and ease of understanding for large repositories. Despite improvements in search algorithms and search benchmarks, the domain of code search has lagged behind. One reason is the high cost of human annotation for code queries and answers. While humans may annotate search results in general text QA systems, code annotations require specialized knowledge of a programming language (PL), as well as domain specific software engineering knowledge. In this work we study the use of Large Language Models (LLMs) to retrieve code at the level of functions and to generate annotations for code search results. We compare the impact of the retriever representation (sparse vs. semantic), programming language, and LLM by comparing human annotations across several popular languages (C, Java, Javascript, Go, and Python). We focus on repositories that implement common data structures likely to be implemented in any PLs. For the same human annotations, we compare several LLM-as-a-Judge models to evaluate programming language and other affinities between LLMs. We find that the chosen retriever and PL exhibit affinities that can be leveraged to improve alignment of human and AI relevance determinations, with significant performance implications. We also find differences in representation (sparse vs. semantic) across PLs that impact alignment of human and AI relevance determinations. We propose using transpilers to bootstrap scalable code search benchmark datasets in other PLs and in a case study demonstrate that human-AI relevance agreement rates largely match the (worst case) human-human agreement under study. The application code used in this work is available at \href{https://github.com/rlucas7/code-searcher/}{this github repo}.

The exponential growth of internet generated data has fueled advancements in artificial intelligence (AI), machine learning (ML), and deep learning (DL) for extracting actionable insights in marketing,telecom, and health sectors. This chapter explores ML applications across three domains namely healthcare, marketing, and telecommunications, with a primary focus on developing a framework for optimal ML algorithm selection. In healthcare, the framework addresses critical challenges such as cardiovascular disease prediction accounting for 28.1% of global deaths and fetal health classification into healthy or unhealthy states, utilizing three datasets. ML algorithms are categorized into eager, lazy, and hybrid learners, selected based on dataset attributes, performance metrics (accuracy, precision, recall), and Akaike Information Criterion (AIC) scores. For validation, eight datasets from the three sectors are employed in the experiments. The key contribution is a recommendation framework that identifies the best ML model according to input attributes, balancing performance evaluation and model complexity to enhance efficiency and accuracy in diverse real-world applications. This approach bridges gaps in automated model selection, offering practical implications for interdisciplinary ML deployment.

Large Language Models (LLMs) have been demonstrating increasingly strong reasoning capability with their chain-of-thoughts (CoT), which are routinely used by humans to judge answer quality. This reliance creates a powerful yet fragile basis for trust. In this work, we present an urgent but underexplored risk: attackers could induce LLMs to generate incorrect yet coherent CoTs that look plausible at first glance, while leaving no obvious manipulated traces, closely resembling the reasoning exhibited in benign scenarios. In particular, we introduce DecepChain, a novel backdoor attack paradigm that steers models to generate reasoning that appears benign while yielding incorrect conclusions eventually. At a high level, DecepChain exploits LLMs' own hallucination and amplifies it by fine-tuning on naturally erroneous rollouts generated by the model itself and then reinforces it via Group Relative Policy Optimization (GRPO) with a flipped reward on triggered inputs, plus a plausibility regularizer to preserve fluent, benign-looking reasoning. Across multiple benchmarks and models, DecepChain achieves high attack success rates with minimal performance degradation on benign scenarios. Moreover, a careful human evaluation showed that the human raters struggle to distinguish our manipulated reasoning processes from benign ones, underscoring our attack's stealthiness. Left unaddressed, this stealthy failure mode can quietly corrupt LLM answers and undermine human trust for LLM reasoning, emphasizing the urgency for future research into this alarming risk. Project page: https://decepchain.github.io/.