The growing ubiquity of conversational AI highlights the need for frameworks that capture not only users' instrumental goals but also the situated, adaptive, and social practices through which they achieve them. Existing taxonomies of conversational behavior either overgeneralize, remain domain-specific, or reduce interactions to narrow dialogue functions. To address this gap, we introduce the Taxonomy of User Needs and Actions (TUNA), an empirically grounded framework developed through iterative qualitative analysis of 1193 human-AI conversations, supplemented by theoretical review and validation across diverse contexts. TUNA organizes user actions into a three-level hierarchy encompassing behaviors associated with information seeking, synthesis, procedural guidance, content creation, social interaction, and meta-conversation. By centering user agency and appropriation practices, TUNA enables multi-scale evaluation, supports policy harmonization across products, and provides a backbone for layering domain-specific taxonomies. This work contributes a systematic vocabulary for describing AI use, advancing both scholarly understanding and practical design of safer, more responsive, and more accountable conversational systems.

Multimodal large language models (MLLMs) have demonstrated remarkable capabilities across vision-language tasks, yet their large-scale deployment raises pressing concerns about memorized private data, outdated knowledge, and harmful content. Existing unlearning approaches for MLLMs typically adapt training-based strategies such as gradient ascent or preference optimization, but these methods are computationally expensive, irreversible, and often distort retained knowledge. In this work, we propose MLLMEraser, an input-aware, training-free framework for test-time unlearning. Our approach leverages activation steering to enable dynamic knowledge erasure without parameter updates. Specifically, we construct a multimodal erasure direction by contrasting adversarially perturbed, knowledge-recall image-text pairs with knowledge-erasure counterparts, capturing both textual and visual discrepancies. To prevent unnecessary interference, we further design an input-aware steering mechanism that adaptively determines when and how the erasure direction should be applied, preserving utility on retained knowledge while enforcing forgetting on designated content.

The pursuit of computational efficiency has driven the adoption of low-precision formats for training transformer models. However, this progress is often hindered by notorious training instabilities. This paper provides the first mechanistic explanation for a long-standing and unresolved failure case where training with flash attention in low-precision settings leads to catastrophic loss explosion. Our in-depth analysis reveals that the failure is not a random artifact but caused by two intertwined phenomena: the emergence of similar low-rank representations within the attention mechanism and the compounding effect of biased rounding errors inherent in low-precision arithmetic. We demonstrate how these factors create a vicious cycle of error accumulation that corrupts weight updates, ultimately derailing the training dynamics. To validate our findings, we introduce a minimal modification to the flash attention that mitigates the bias in rounding errors. This simple change stabilizes the training process, confirming our analysis and offering a practical solution to this persistent problem. The pursuit of training ever-larger and more powerful transformer models is a relentless drive for computational efficiency (Brown et al., 2020; Hoffmann et al., 2022). A key strategy in this endeavor is the adoption of low-precision numerical formats (Micikevicius et al., 2017; Wang et al., 2018; Kalamkar et al., 2019; Liu et al., 2024), which promise substantial reductions in memory footprint and significant boosts in training speed. In industrial practice, it is common to use BF16 for memory-bound operations like flash attention while pushing compute-bound operations like FFNs to even lower precisions such as FP8 (Liu et al., 2024; Qwen-Team, 2025). This highlights the heightened sensitivity of attention mechanisms to numerical precision.

We release Code World Model (CWM), a 32-billion-parameter open-weights LLM, to advance research on code generation with world models. To improve code understanding beyond what can be learned from training on static code alone, we mid-train CWM on a large amount of observation-action trajectories from Python interpreter and agentic Docker environments, and perform extensive multi-task reasoning RL in verifiable coding, math, and multi-turn software engineering environments. With CWM, we provide a strong testbed for researchers to explore the opportunities world modeling affords for improving code generation with reasoning and planning in computational environments. We present first steps of how world models can benefit agentic coding, enable step-by-step simulation of Python code execution, and show early results of how reasoning can benefit from the latter. CWM is a dense, decoder-only LLM trained with a context size of up to 131k tokens. Independent of its world modeling capabilities, CWM offers strong performance on general coding and math tasks: it reaches pass@1 scores of 65.8% on SWE-bench Verified (with test-time scaling), 68.6% on LiveCodeBench, 96.6% on Math-500, and 76.0% on AIME 2024. To support further research on code world modeling, we release model checkpoints after mid-training, SFT, and RL.

Simulating in silico cellular responses to interventions is a promising direction to accelerate high-content image-based assays, critical for advancing drug discovery and gene editing. To support this, we introduce MorphGen, a state-of-the-art diffusion-based generative model for fluorescent microscopy that enables controllable generation across multiple cell types and perturbations. To capture biologically meaningful patterns consistent with known cellular morphologies, MorphGen is trained with an alignment loss to match its representations to the phenotypic embeddings of OpenPhenom, a state-of-the-art biological foundation model. Unlike prior approaches that compress multichannel stains into RGB images -- thus sacrificing organelle-specific detail -- MorphGen generates the complete set of fluorescent channels jointly, preserving per-organelle structures and enabling a fine-grained morphological analysis that is essential for biological interpretation. We demonstrate biological consistency with real images via CellProfiler features, and MorphGen attains an FID score over 35% lower than the prior state-of-the-art MorphoDiff, which only generates RGB images for a single cell type. Code is available at https://github.com/czi-ai/MorphGen.

We introduce ClustRecNet - a novel deep learning (DL)-based recommendation framework for determining the most suitable clustering algorithms for a given dataset, addressing the long-standing challenge of clustering algorithm selection in unsupervised learning. To enable supervised learning in this context, we construct a comprehensive data repository comprising 34,000 synthetic datasets with diverse structural properties. Each of them was processed using 10 popular clustering algorithms. The resulting clusterings were assessed via the Adjusted Rand Index (ARI) to establish ground truth labels, used for training and evaluation of our DL model. The proposed network architecture integrates convolutional, residual, and attention mechanisms to capture both local and global structural patterns from the input data. This design supports end-to-end training to learn compact representations of datasets and enables direct recommendation of the most suitable clustering algorithm, reducing reliance on handcrafted meta-features and traditional Cluster Validity Indices (CVIs). Comprehensive experiments across synthetic and real-world benchmarks demonstrate that our DL model consistently outperforms conventional CVIs (e.g. Silhouette, Calinski-Harabasz, Davies-Bouldin, and Dunn) as well as state-of-the-art AutoML clustering recommendation approaches (e.g. ML2DAC, AutoCluster, and AutoML4Clust). Notably, the proposed model achieves a 0.497 ARI improvement over the Calinski-Harabasz index on synthetic data and a 15.3% ARI gain over the best-performing AutoML approach on real-world data.

Single-cell RNA sequencing has transformed biology by enabling the measurement of gene expression at cellular resolution, providing information for cell types, states, and disease contexts. Recently, single-cell foundation models have emerged as powerful tools for learning transferable representations directly from expression profiles, improving performance on classification and clustering tasks. However, these models are limited to discrete prediction heads, which collapse cellular complexity into predefined labels that fail to capture the richer, contextual explanations biologists need. We introduce Cell2Text, a multimodal generative framework that translates scRNA-seq profiles into structured natural language descriptions. By integrating gene-level embeddings from single-cell foundation models with pretrained large language models, Cell2Text generates coherent summaries that capture cellular identity, tissue origin, disease associations, and pathway activity, generalizing to unseen cells. Empirically, Cell2Text outperforms baselines on classification accuracy, demonstrates strong ontological consistency using PageRank-based similarity metrics, and achieves high semantic fidelity in text generation. These results demonstrate that coupling expression data with natural language offers both stronger predictive performance and inherently interpretable outputs, pointing to a scalable path for label-efficient characterization of unseen cells.

LiDAR's dense, sharp point cloud (PC) representations of the surrounding environment enable accurate perception and significantly improve road safety by offering greater scene awareness and understanding. However, LiDAR's high cost continues to restrict the broad adoption of high-level Autonomous Driving (AD) systems in commercially available vehicles. Prior research has shown progress towards circumventing the need for LiDAR by training a neural network, using LiDAR point clouds as ground truth (GT), to produce LiDAR-like 3D point clouds using only 4D Radars. One of the best examples is a neural network created to train a more efficient radar target detector with a modular 2D convolutional neural network (CNN) backbone and a temporal coherence network at its core that uses the RaDelft dataset for training (see arXiv:2406.04723). In this work, we investigate the impact of higher-capacity segmentation backbones on the quality of the produced point clouds. Our results show that while very high-capacity models may actually hurt performance, an optimal segmentation backbone can provide a 23.7% improvement over the state-of-the-art (SOTA).

Making large language models (LLMs) more efficient in memory, latency, and serving cost is crucial for edge deployment, interactive applications, and sustainable inference at scale. Pruning is a promising technique, but existing pruning methods are limited: width pruning often breaks the standard transformer layout, requiring custom inference code, while depth pruning can cause abrupt accuracy drops. Also, while many pruning approaches are effective against LLMs, they struggle to maintain performance on small language models (SLMs). In this work, we propose COMPACT, which jointly (i) prunes rare vocabulary to shrink embedding/LM head layers and (ii) prunes FFN intermediate channels using common-token-weighted activations, aligning importance with the post-pruning token distribution. COMPACT inherits strengths of both depth and width pruning, such as: deployment-friendliness (keeps a standard transformer architecture), scale-adaptivity (trade off vocab. vs. FFN pruning), competitive pruning times, and strong memory savings alongside throughput gains. Experiments across Qwen, LLaMA, and Gemma families (0.5B-70B) show state-of-the-art downstream performance, with substantial reductions in parameters, GPU memory, and latency.

Modern recurrent deep learning models - such as state-space models (SSMs) - have emerged as a promising computationally efficient alternative to Transformers for sequence modeling. However, how their practical differences in learn-ability and optimization impact core capabilities remains underexplored. In this paper, we thoroughly compare SSM and Transformer learning dynamics on two fundamental benchmarks highly correlated with language modeling performance: associative recall and copying. We find that, while Transformers are robust to optimization hyperparameters, the performance of modern recurrent models suffers from critical instabilities: success is confined to an extremely narrow window of learning rates, outside of which accuracy drastically drops. This issue can confound performance evaluations and expressivity conclusions, revealing a fundamental mismatch in the loss landscape of modern recurrent models compared to Transformers. We demonstrate that this brittle optimization has a direct impact on scaling, causing SSMs to favor width over depth. Indeed, we also find that, while the 1-layer Transformer's performance on recall does not exceed random guessing, well-tuned Mamba and other SSMs can learn to recall with one layer, yet with dynamics that do not resemble the formation of induction heads. Taken together, our findings suggest that a crucial differentiator between these architectures lies not just in their expressivity but in their fundamental learnability properties, pointing to optimization stability as a key challenge for the future of SSMs. Since early developments (Rumelhart et al., 1986; Elman, 1990), RNNs have driven progress in machine learning techniques for sequential data, with milestones such as Echo-State Networks (Jaeger, 2001) LSTM (Hochreiter & Schmidhuber, 1997) and GRU (Cho et al., 2014). However, two problems severely limit the application of RNNs in modern times: first, GPU architectures struggle with sequential processing. Secondly, it is widely known that RNNs are hard to train due to vanishing and exploding gradients issues (Bengio et al., 1994; Hochreiter et al., 2001; Pascanu et al., 2013). These challenges have led to the introduction of a different paradigm: the Attention mechanism, implemented around the Transformer architecture (V aswani et al., 2017). Instead of processing inputs sequentially while building up internal memory (RNNs), Attention computes pairwise interactions between data points, allowing for modeling direct links between elements in a sequence and thus mitigating vanishing gradients.