Data science is a pillar of artificial intelligence (AI), which is transforming nearly every domain of human activity, from the social and physical sciences to engineering and medicine. While data-driven findings in AI offer unprecedented power to extract insights and guide decision-making, many are difficult or impossible to replicate. A key reason for this challenge is the uncertainty introduced by the many choices made throughout the data science life cycle (DSLC). Traditional statistical frameworks often fail to account for this uncertainty. The Predictability-Computability-Stability (PCS) framework for veridical (truthful) data science offers a principled approach to addressing this challenge throughout the DSLC. This paper presents an updated and streamlined PCS workflow, tailored for practitioners and enhanced with guided use of generative AI. We include a running example to display the PCS framework in action, and conduct a related case study which showcases the uncertainty in downstream predictions caused by judgment calls in the data cleaning stage.

Reinforcement learning (RL) is a critical component of large language model (LLM) post-training. However, on-policy algorithms used for post-training are not naturally robust to a diversified content of experience replay buffers, which asynchronous off-policy actors can efficiently populate in parallel to training. We propose efficiently learning on such off-policy data via Trajectory Balance with Asynchrony (TBA), an approach to asynchronous RL for LLMs that leverages the principled off-policy TB objective. On math, preference-tuning, and automated red-teaming tasks, we post-train models ranging from Pythia 410M to Qwen 2.5 7B, finding TBA offers speed and performance boosts over strong baselines like Online DPO and Dr. GRPO. Beyond TBA's performance benefits (high accuracy even as asynchrony grows) and speedups ($4\times$ or more), we show its reward- and recency-prioritizing sampling enable further gains as data generation is scaled. Our code is available at https://github.com/bbartoldson/TBA.

Vision Language Models (VLMs) demonstrate strong qualitative visual understanding, but struggle with metrically precise spatial reasoning required for embodied applications. The agentic paradigm promises that VLMs can use a wide variety of tools that could augment these capabilities, such as depth estimators, segmentation models, and pose estimators. Yet it remains an open challenge how to realize this vision without solely relying on handcrafted prompting strategies or enforcing fixed, predefined tool pipelines that limit VLMs' ability to discover optimal tool-use patterns. Reinforcement Learning could overcome this gap, but has so far been limited to reasoning with a single visual tool due to the large search space in multi-tool reasoning. We introduce Double Interactive Reinforcement Learning (DIRL), a two-phase training framework where VLMs learn to coordinate multiple tools through interactive exploration and feedback. In the teaching phase, we combine demonstrations from a single tool specialist trained via interactive RL with traces from a frontier model using all tools. In the exploration phase, the model further refines multi-tool coordination through continued RL. Our model, SpaceTools, with tool-augmented spatial reasoning ability, achieves state-of-the-art performance on spatial understanding benchmarks (RoboSpatial-Home, BLINK, BOP-ASK) and demonstrates reliable real-world manipulation using a 7-DOF robot as a tool. DIRL provides substantial improvements over the vanilla SFT (+12% on RoboSpatial) and RL (+16% on RoboSpatial) baselines. Project page: https://spacetools.github.io/.

Watermarking aims to embed hidden signals in generated text that can be reliably detected when given access to a secret key. Open-weight language models pose acute challenges for such watermarking schemes because the inference-time interventions that dominate contemporary approaches cannot be enforced once model weights are public. Existing watermaking techniques for open-weight models, such as the recently proposed GaussMark, typically rely on small modifications to model weights, which can yield signals detectable to those equipped with a secret key, but achieving detection power comparable to inference-time watermarks generally requires weight perturbations that noticeably reduce generation quality. We introduce MarkTune, a theoretically principled, on-policy fine-tuning framework that treats the GaussMark signal as a reward while simultaneously regularizing against degradation in text quality. We derive MarkTune as an improvement on GaussMark and demonstrate that MarkTune consistently improves the quality-detectability trade-off over GaussMark by steering finer-grained, watermark-aware weight updates within the model's representation space while preserving generation quality. Empirically, we show that MarkTune pushes the quality-detectability frontier of GaussMark close to that of inference-time watermarking, remains robust to paraphrasing and fine-tuning attacks, and exhibits strong generalization: a model fine-tuned on one dataset retains substantial watermark detection power on unseen datasets. Together, these results establish MarkTune as a general strategy for embedding robust, high-quality watermarks into open-weight LMs.

Why do state-of-the-art OOD detection methods exhibit catastrophic failure when models are trained on single-domain datasets? We provide the first theoretical explanation for this phenomenon through the lens of information theory. We prove that supervised learning on single-domain data inevitably produces domain feature collapse -- representations where I(x_d; z) = 0, meaning domain-specific information is completely discarded. This is a fundamental consequence of information bottleneck optimization: models trained on single domains (e.g., medical images) learn to rely solely on class-specific features while discarding domain features, leading to catastrophic failure when detecting out-of-domain samples (e.g., achieving only 53% FPR@95 on MNIST). We extend our analysis using Fano's inequality to quantify partial collapse in practical scenarios. To validate our theory, we introduce Domain Bench, a benchmark of single-domain datasets, and demonstrate that preserving I(x_d; z) > 0 through domain filtering (using pretrained representations) resolves the failure mode. While domain filtering itself is conceptually straightforward, its effectiveness provides strong empirical evidence for our information-theoretic framework. Our work explains a puzzling empirical phenomenon, reveals fundamental limitations of supervised learning in narrow domains, and has broader implications for transfer learning and when to fine-tune versus freeze pretrained models.

The application of Large Multimodal Models (LMMs) to long-form video understanding is constrained by limited context lengths and the computationally prohibitive cost of processing dense video tokens. Consequently, recent research has focused on query-aware frame selection, methods that often incur significant computational overhead. This paper challenges the assumption that such complex search mechanisms are universally necessary. W e first identify and validate a query typology distinguishing between global query and localized query. W e demonstrate that while uniform sampling is both effective and efficient for global queries, localized queries indeed necessitate query-aware selection for optimal performance. Building on this insight, we propose DIG, a training-free frame selection framework that adapts its strategy based on the query type. Specifically, DIG employs efficient uniform sampling for global queries while activating a specialized pipeline to extract query-relevant frames for localized queries. Experiments on three long-form video understanding benchmarks demonstrate that DIG consistently outperforms existing baselines and robustly improves LMM performance, even when scaling the input frame count to 256.

Vision-Language Models (VLMs) deployed in safety-critical applications such as autonomous driving must handle continuous visual streams under imperfect conditions. However, existing benchmarks focus on static, high-quality images and ignore temporal degradation and error propagation, which are critical failure modes where transient visual corruption induces hallucinations that persist across subsequent frames. We introduce DIQ-H, the first benchmark for evaluating VLM robustness under dynamic visual degradation in temporal sequences. DIQ-H applies physics-based corruptions including motion blur, sensor noise, and compression artifacts, and measures hallucination persistence, error recovery, and temporal consistency through multi-turn question-answering tasks. To enable scalable annotation, we propose Uncertainty-Guided Iterative Refinement (UIR), which generates reliable pseudo-ground-truth using lightweight VLMs with uncertainty filtering, achieving a 15.3 percent accuracy improvement. Experiments on 16 state-of-the-art VLMs reveal substantial robustness gaps: even advanced models such as GPT-4o achieve only a 78.5 percent recovery rate, while open-source models struggle with temporal consistency at less than 60 percent. DIQ-H provides a comprehensive platform for evaluating VLM reliability in real-world deployments.

Adapting large language models (LLMs) to low-resource languages remains a major challenge due to data scarcity and cross-lingual drift. This work presents a two-stage adaptation of Qwen2.5-3B to Tibetan, a morphologically rich and underrepresented language. We employ Continual Pretraining (CPT) to establish Tibetan linguistic grounding, followed by Supervised Fine-Tuning (SFT) for task and translation specialization. Empirical evaluations demonstrate a consistent decrease in perplexity (from 2.98 $\rightarrow$ 1.54) and substantial improvements in Chinese$\rightarrow$Tibetan translation quality (BLEU: 0.046 $\rightarrow$ 0.261; chrF: 2.2 $\rightarrow$ 6.6). Layer-wise analysis across 435 layers in Qwen3-4B reveals that adaptation primarily concentrates on embedding and output heads, with mid--late MLP projections encoding domain-specific transformations. Our findings suggest that CPT constructs a Tibetan semantic manifold while SFT sharpens task alignment with minimal representational disruption. This study provides the first quantitative exploration of Tibetan adaptation dynamics for LLMs, and offers an open, reproducible framework for extending multilingual foundation models to low-resource settings.

In the vision domain, dataset distillation arises as a technique to condense a large dataset into a smaller synthetic one that exhibits a similar result in the training process. While image data presents an extensive literature of distillation methods, text dataset distillation has fewer works in comparison. Text dataset distillation initially grew as an adaptation of efforts from the vision universe, as the particularities of the modality became clear obstacles, it rose into a separate branch of research. Several milestones mark the development of this area, such as the introduction of methods that use transformer models, the generation of discrete synthetic text, and the scaling to decoder-only models with over 1B parameters. Despite major advances in modern approaches, the field remains in a maturing phase, with room for improvement on benchmarking standardization, approaches to overcome the discrete nature of text, handling complex tasks, and providing explicit examples of real-world applications. In this report, we review past and recent advances in dataset distillation for text, highlighting different distillation strategies, key contributions, and general challenges.

While recent developments in large language models have improved bias detection and classification, sensitive subjects like religion still present challenges because even minor errors can result in severe misunderstandings. In particular, multilingual models often misrepresent religions and have difficulties being accurate in religious contexts. To address this, we introduce BRAND: Bilingual Religious Accountable Norm Dataset, which focuses on the four main religions of South Asia: Buddhism, Christianity, Hinduism, and Islam, containing over 2,400 entries, and we used three different types of prompts in both English and Bengali. Our results indicate that models perform better in English than in Bengali and consistently display bias toward Islam, even when answering religion-neutral questions. These findings highlight persistent bias in multilingual models when similar questions are asked in different languages. We further connect our findings to the broader issues in HCI regarding religion and spirituality.