Kevin Lee

Gheorghe Comanici, Eric Bieber, Mike Schaekermann et al. · amazon-science, baidu

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Kevin Lee, Pablo Millan Arias

Layer-wise Relevance Propagation (LRP) provides principled attribution for neural networks through conservation properties and foundations in Deep Taylor Decomposition. However, existing implementations operate at the module level, requiring architecture-specific propagation rules and modifications. These limit the generality of target model and sustainability of implementations as architectures evolve. We introduce DynamicLRP, a model-agnostic LRP framework operating at the tensor operation level. By decomposing attribution to individual operations within computation graphs and introducing a novel mechanism for deferred activation resolution, named the Promise System, our approach achieves true architecture agnosticity while maintaining LRP's theoretical guarantees. This design operates independently of backpropagation machinery, enabling operation on arbitrary computation graphs without model modification and side-by-side execution with gradient backpropagation. Being based on computation graphs, this method is theoretically extensible to other deep learning libraries that support auto-differentiation. We demonstrate faithfulness matching or exceeding specialized implementations (1.77 vs 1.69 ABPC on VGG, equivalent performance on ViT, 93.70\% and 95.06\% top-1 attribution accuracy for explaining RoBERTa-large and Flan-T5-large answers on SQuADv2, respectively) while maintaining practical efficiency on models with hundreds of millions of parameters. We achieved 99.92\% node coverage across 31,465 computation graph nodes from 15 diverse architectures, including state-space models (Mamba), audio transformers (Whisper), and multimodal systems (DePlot) without any model-specific code with rules for 47 fundamental operations implemented. Our operation-level decomposition and Promise System establish a sustainable, extensible foundation for LRP across evolving architectures.

Otto Brookes, Maksim Kukushkin, Majid Mirmehdi et al.

Computer vision analysis of camera trap video footage is essential for wildlife conservation, as captured behaviours offer some of the earliest indicators of changes in population health. Recently, several high-impact animal behaviour datasets and methods have been introduced to encourage their use; however, the role of behaviour-correlated background information and its significant effect on out-of-distribution generalisation remain unexplored. In response, we present the PanAf-FGBG dataset, featuring 20 hours of wild chimpanzee behaviours, recorded at over 350 individual camera locations. Uniquely, it pairs every video with a chimpanzee (referred to as a foreground video) with a corresponding background video (with no chimpanzee) from the same camera location. We present two views of the dataset: one with overlapping camera locations and one with disjoint locations. This setup enables, for the first time, direct evaluation of in-distribution and out-of-distribution conditions, and for the impact of backgrounds on behaviour recognition models to be quantified. All clips come with rich behavioural annotations and metadata including unique camera IDs and detailed textual scene descriptions. Additionally, we establish several baselines and present a highly effective latent-space normalisation technique that boosts out-of-distribution performance by +5.42% mAP for convolutional and +3.75% mAP for transformer-based models. Finally, we provide an in-depth analysis on the role of backgrounds in out-of-distribution behaviour recognition, including the so far unexplored impact of background durations (i.e., the count of background frames within foreground videos).

Vijay Prakash, Kevin Lee, Arkaprabha Bhattacharya et al.

Answering end user security questions is challenging. While large language models (LLMs) like GPT, LLAMA, and Gemini are far from error-free, they have shown promise in answering a variety of questions outside of security. We studied LLM performance in the area of end user security by qualitatively evaluating 3 popular LLMs on 900 systematically collected end user security questions. While LLMs demonstrate broad generalist ``knowledge'' of end user security information, there are patterns of errors and limitations across LLMs consisting of stale and inaccurate answers, and indirect or unresponsive communication styles, all of which impacts the quality of information received. Based on these patterns, we suggest directions for model improvement and recommend user strategies for interacting with LLMs when seeking assistance with security.

Kevin Lee, Russell Spiewak, James Walsh

Scientific reasoning through Large Language Models in heliophysics involves more than just recalling facts: it requires incorporating physical assumptions, maintaining consistent units, and providing clear scientific formats through coordinated approaches. To address these challenges, we present Reasoning With a Star, a newly contributed heliophysics dataset applicable to reasoning; we also provide an initial benchmarking approach. Our data are constructed from National Aeronautics and Space Administration & University Corporation for Atmospheric Research Living With a Star summer school problem sets and compiled into a readily consumable question-and-answer structure with question contexts, reasoning steps, expected answer type, ground-truth targets, format hints, and metadata. A programmatic grader checks the predictions using unit-aware numerical tolerance, symbolic equivalence, and schema validation. We benchmark a single-shot baseline and four multi-agent patterns, finding that decomposing workflows through systems engineering principles outperforms direct prompting on problems requiring deductive reasoning rather than pure inductive recall.

Kevin Lee

NOAA's Deep-ocean Assessment and Reporting of Tsunamis (DART) data are critical for NASA-JPL's tsunami detection, real-time operations, and oceanographic research. However, these time-series data often contain spikes, steps, and drifts that degrade data quality and obscure essential oceanographic features. To address these anomalies, the work introduces an Iterative Encoding-Decoding Variational Autoencoders (Iterative Encoding-Decoding VAEs) model to improve the quality of DART time series. Unlike traditional filtering and thresholding methods that risk distorting inherent signal characteristics, Iterative Encoding-Decoding VAEs progressively remove anomalies while preserving the data's latent structure. A hybrid thresholding approach further retains genuine oceanographic features near boundaries. Applied to complex DART datasets, this approach yields reconstructions that better maintain key oceanic properties compared to classical statistical techniques, offering improved robustness against spike removal and subtle step changes. The resulting high-quality data supports critical verification and validation efforts for the GRACE-FO mission at NASA-JPL, where accurate surface measurements are essential to modeling Earth's gravitational field and global water dynamics. Ultimately, this data processing method enhances tsunami detection and underpins future climate modeling with improved interpretability and reliability.

Otto Brookes, Majid Mirmehdi, Colleen Stephens et al.

We present the PanAf20K dataset, the largest and most diverse open-access annotated video dataset of great apes in their natural environment. It comprises more than 7 million frames across ~20,000 camera trap videos of chimpanzees and gorillas collected at 14 field sites in tropical Africa as part of the Pan African Programme: The Cultured Chimpanzee. The footage is accompanied by a rich set of annotations and benchmarks making it suitable for training and testing a variety of challenging and ecologically important computer vision tasks including ape detection and behaviour recognition. Furthering AI analysis of camera trap information is critical given the International Union for Conservation of Nature now lists all species in the great ape family as either Endangered or Critically Endangered. We hope the dataset can form a solid basis for engagement of the AI community to improve performance, efficiency, and result interpretation in order to support assessments of great ape presence, abundance, distribution, and behaviour and thereby aid conservation efforts.

Ngoc Dung Huynh, Mohamed Reda Bouadjenek, Imran Razzak et al.

A Machine-Critical Application is a system that is fundamentally necessary to the success of specific and sensitive operations such as search and recovery, rescue, military, and emergency management actions. Recent advances in Machine Learning, Natural Language Processing, voice recognition, and speech processing technologies have naturally allowed the development and deployment of speech-based conversational interfaces to interact with various machine-critical applications. While these conversational interfaces have allowed users to give voice commands to carry out strategic and critical activities, their robustness to adversarial attacks remains uncertain and unclear. Indeed, Adversarial Artificial Intelligence (AI) which refers to a set of techniques that attempt to fool machine learning models with deceptive data, is a growing threat in the AI and machine learning research community, in particular for machine-critical applications. The most common reason of adversarial attacks is to cause a malfunction in a machine learning model. An adversarial attack might entail presenting a model with inaccurate or fabricated samples as it's training data, or introducing maliciously designed data to deceive an already trained model. While focusing on speech recognition for machine-critical applications, in this paper, we first review existing speech recognition techniques, then, we investigate the effectiveness of adversarial attacks and defenses against these systems, before outlining research challenges, defense recommendations, and future work. This paper is expected to serve researchers and practitioners as a reference to help them in understanding the challenges, position themselves and, ultimately, help them to improve existing models of speech recognition for mission-critical applications. Keywords: Mission-Critical Applications, Adversarial AI, Speech Recognition Systems.

Ben Kaiser, Jerry Wei, Eli Lucherini et al.

Disinformation is proliferating on the internet, and platforms are responding by attaching warnings to content. There is little evidence, however, that these warnings help users identify or avoid disinformation. In this work, we adapt methods and results from the information security warning literature in order to design and evaluate effective disinformation warnings. In an initial laboratory study, we used a simulated search task to examine contextual and interstitial disinformation warning designs. We found that users routinely ignore contextual warnings, but users notice interstitial warnings -- and respond by seeking information from alternative sources. We then conducted a follow-on crowdworker study with eight interstitial warning designs. We confirmed a significant impact on user information-seeking behavior, and we found that a warning's design could effectively inform users or convey a risk of harm. We also found, however, that neither user comprehension nor fear of harm moderated behavioral effects. Our work provides evidence that disinformation warnings can -- when designed well -- help users identify and avoid disinformation. We show a path forward for designing effective warnings, and we contribute repeatable methods for evaluating behavioral effects. We also surface a possible dilemma: disinformation warnings might be able to inform users and guide behavior, but the behavioral effects might result from user experience friction, not informed decision making.

Casey Casalnuovo, Kevin Lee, Hulin Wang et al.

Natural code is known to be very repetitive (much more so than natural language corpora); furthermore, this repetitiveness persists, even after accounting for the simpler syntax of code. However, programming languages are very expressive, allowing a great many different ways (all clear and unambiguous) to express even very simple computations. So why is natural code repetitive? We hypothesize that the reasons for this lie in fact that code is bimodal: it is executed by machines, but also read by humans. This bimodality, we argue, leads developers to write code in certain preferred ways that would be familiar to code readers. To test this theory, we 1) model familiarity using a language model estimated over a large training corpus and 2) run an experiment applying several meaning preserving transformations to Java and Python expressions in a distinct test corpus to see if forms more familiar to readers (as predicted by the language models) are in fact the ones actually written. We find that these transformations generally produce program structures that are less common in practice, supporting the theory that the high repetitiveness in code is a matter of deliberate preference. Finally, 3) we use a human subject study to show alignment between language model score and human preference for the first time in code, providing support for using this measure to improve code.

Jason M. Gregory, Christopher Reardon, Kevin Lee et al.

Human-robot teaming offers great potential because of the opportunities to combine strengths of heterogeneous agents. However, one of the critical challenges in realizing an effective human-robot team is efficient information exchange - both from the human to the robot as well as from the robot to the human. In this work, we present and analyze an augmented reality-enabled, gesture-based system that supports intuitive human-robot teaming through improved information exchange. Our proposed system requires no external instrumentation aside from human-wearable devices and shows promise of real-world applicability for service-oriented missions. Additionally, we present preliminary results from a pilot study with human participants, and highlight lessons learned and open research questions that may help direct future development, fielding, and experimentation of autonomous HRI systems.

Malte Möser, Kyle Soska, Ethan Heilman et al.

Monero is a privacy-centric cryptocurrency that allows users to obscure their transactions by including chaff coins, called "mixins," along with the actual coins they spend. In this paper, we empirically evaluate two weaknesses in Monero's mixin sampling strategy. First, about 62% of transaction inputs with one or more mixins are vulnerable to "chain-reaction" analysis -- that is, the real input can be deduced by elimination. Second, Monero mixins are sampled in such a way that they can be easily distinguished from the real coins by their age distribution; in short, the real input is usually the "newest" input. We estimate that this heuristic can be used to guess the real input with 80% accuracy over all transactions with 1 or more mixins. Next, we turn to the Monero ecosystem and study the importance of mining pools and the former anonymous marketplace AlphaBay on the transaction volume. We find that after removing mining pool activity, there remains a large amount of potentially privacy-sensitive transactions that are affected by these weaknesses. We propose and evaluate two countermeasures that can improve the privacy of future transactions.

Kevin Lee, Lingzhou Xue

Graphical model has been widely used to investigate the complex dependence structure of high-dimensional data, and it is common to assume that observed data follow a homogeneous graphical model. However, observations usually come from different resources and have heterogeneous hidden commonality in real-world applications. Thus, it is of great importance to estimate heterogeneous dependencies and discover subpopulation with certain commonality across the whole population. In this work, we introduce a novel regularized estimation scheme for learning nonparametric mixture of Gaussian graphical models, which extends the methodology and applicability of Gaussian graphical models and mixture models. We propose a unified penalized likelihood approach to effectively estimate nonparametric functional parameters and heterogeneous graphical parameters. We further design an efficient generalized effective EM algorithm to address three significant challenges: high-dimensionality, non-convexity, and label switching. Theoretically, we study both the algorithmic convergence of our proposed algorithm and the asymptotic properties of our proposed estimators. Numerically, we demonstrate the performance of our method in simulation studies and a real application to estimate human brain functional connectivity from ADHD imaging data, where two heterogeneous conditional dependencies are explained through profiling demographic variables and supported by existing scientific findings.