Category
490 papers
These findings show that reliable evaluation of LLM-generated code requires validated ground truth, protected tests, and multiple explicitly interpreted measures, and that CodeAssay provides a reproducible basis for evidence-based model evaluation in AI-augmented software development.
LookBack, a training-free LVLM response scoring method that augments token likelihood with visual lookback score, a lightweight measure of how strongly each response token refers to image tokens, consistently improves Best-of-$N$ selection over existing baselines with negligible additional overhead.
Current structure-based drug design generative models often struggle to faithfully recapitulate genuine ligand-protein binding interactions. Instead, under the coupling of implicit learning architectures and biased training data, they tend to learn spurious statistical correlations. To address this, we propose EIP-Diff (Explicit Interaction-Prompted Diffusion), an architecture featuring a novel explicit interaction-prompt embedding mechanism that is better suited for real-world target-specific drug design. This architecture replaces biased implicit learning with explicit, residue-level biological guidance, thereby promoting more fine-grained geometric fidelity and more precise interaction-aware conditioning. To fully realize the capabilities of EIP-Diff and provide a reliable basis for performance evaluation, we further constructed CrystalData set, which provides higher-fidelity and less-biased structural supervision than existing data sets. This explicit architecture markedly improves distribution consistency: even when trained on the crossdocked data set, EIP-Diff achieves the highest alignment with authentic pharmacological distributions among evaluated models. Training on CrystalData set further enhances this alignment and improves 3D geometric accuracy, while retaining strong controllability, high chemical space coverage, and near-perfect uniqueness. In addition, target-based validation on KAT6A and YTHDC1 confirmed that EIP-Diff accurately recapitulates native-like binding modes. Furthermore, in a real-world drug design task against IDO1, we successfully designed a novel lead compound with nanomolar potency (IC50 = 0.31 nM). These results demonstrate that the EIP-Diff architecture can explicitly leverage experimentally derived structural data and biologically meaningful interaction information for target-specific molecular generation, thereby enabling its effective application to real-world structure-based drug design.
Post-training alignment is often shallow, eroding under fine-tuning. It remains untested as to whether constitutional midtraining interventions can produce durable alignment when cleanly isolated from post-training. We build a 394M-token constitutional corpus from Anthropic's Constitution and apply constitutional midtraining at 120B scale, where principled, values-based content is inserted into midtraining. A 2x2 design (curriculum ordering x deliberative reasoning) was used to produce four constitutionally midtrained conditions, plus a control, which were evaluated on self-generated and established benchmarks including alignment under pressure, value conflict resolution, blackmail, and emergent misalignment. All models were evaluated across three stages: post-midtraining, post-SFT, and post-benign fine-tuning. Constitutionally midtrained models outperformed the control on alignment generalization and durability, notably on blackmail: SFT instilled a blackmail propensity in all models, but constitutional midtraining blunted it, with the advantage surviving benign fine-tuning (-17.5pp). This durability did not extend to settings that required active resistance to in-context pressure or conflict, where the advantage attenuates after SFT. The presence of constitutional content at midtraining also mattered more than its structure, and constitutional midtraining incurred no capability cost, on average, at any stage (MMLU, ARC-Easy, piqa, GSM8K). A modest amount of constitutional content at midtraining could therefore yield broad, persistent alignment gains, offering a cheap, complementary addition to SFT-centered pipelines. Code, data, and models are available.
Reach audiences
Advertise in front of researchers, engineers, and readers.
This work investigates MRI-based Microbial Density Stratification as a patient-level representation learning task, and Center Heatmap-driven Macro-micro modeling Network (CHM-Net) is introduced for this task, establishing the link between imaging phenotypes and microbial states through center heatmap-guided small-lesion response localization.
A lightweight debiasing framework for pretrained transformer-based encoders that improves prediction fairness while preserving task performance and reduces sensitive-attribute attacker accuracy, improves fairness metrics such as demographic parity difference and equalized odds, and maintains competitive task performance.
SegWithU is a post-hoc framework that augments a frozen pretrained segmentation backbone with a lightweight uncertainty head and models uncertainty as perturbation energy in a compact probe space using rank-1 posterior probes, suggesting that perturbation-based uncertainty modeling is an effective and practical route to reliability-aware medical segmentation.
Comparing diffusion and rectified flow objectives within the MotionGPT3 framework suggests that several known benefits of rectified flow objectives do extend to continuous-latent text-to-motion generation, highlighting the importance of the training objective choice in motion priors.
This work introduces a pioneering exploration of Self-Supervised Learning (SSL) within the SNN, and proposes a novel Spiking Self-Attention (SSA) and Spiking Transformer (Spikformer) that achieves 80+% accuracy on ImageNet.
This study addresses critical gaps in automated lymphoma segmentation from PET/CT images, focusing on issues often overlooked in existing literature. While deep learning has been applied for lymphoma lesion segmentation, few studies incorporate out-of-distribution testing, raising concerns about model generalizability across diverse imaging conditions and patient populations. We highlight the need to compare model performance with expert human annotators, including intra- and inter-observer variability, to understand task difficulty better. Most approaches focus on overall segmentation accuracy but overlook lesion-specific measures important for precise lesion detection and disease quantification. To address these gaps, we propose a clinically relevant framework for evaluating deep segmentation networks. Using this lesion measure-specific evaluation, we assess the performance of four deep networks (ResUNet, SegResNet, DynUNet, and SwinUNETR) across 611 cases from multi-institutional datasets, covering various lymphoma subtypes and lesion characteristics. Beyond standard metrics like the Dice similarity coefficient, we evaluate clinical lesion measures and their prediction errors. We also introduce detection criteria for lesion localization and propose a new detection Criterion 3 based on metabolic characteristics. We show that networks perform better on large, intense lesions with higher metabolic activity. Finally, we compare network performance to physicians via intra- and inter-observer variability analyses, demonstrating that network errors closely resemble those made by experts, i.e., the small and faint lesions remain challenging for both humans and networks. This study aims to improve automated lesion segmentation's clinical relevance, supporting better treatment decisions for lymphoma patients. The code is available at: https://github.com/microsoft/lymphoma-segmentation-dnn.
OceanDepths is introduced, the first open, global, regridded AI-ready dataset that pairs satellite-derived sea surface temperature, sea surface salinity, and sea surface height products with co-located EN4 subsurface temperature and salinity profiles, complemented by matched GLORYS12 ocean reanalysis data to support comparisons or multi-stage learning.
Can neural networks learn algebraic rules from visual inputs, or do they merely fit observed patterns? We study this question using MNIST (or EMNIST letters) as states and modular arithmetic operations as actions in a JEPA-style world model. Standard supervised and JEPA baselines with operation embeddings achieve high accuracy on seen operations but fail to extrapolate reliably to unseen operations. We propose BRo-JEPA, a world model with a block-rotation predictor that represents arithmetic operations as rotations, resulting in the cyclic structure of modular arithmetic in latent space. By applying actions as rotations, the BRo-JEPA predictor learns the rotation angles to align the latent representations with the underlying modular structure which enables strict zero-shot operation generalization. While our best block-rotation supervised baseline reaches only 54.54% zero-shot accuracy on MNIST and 25.13% on EMNIST, BRo-JEPA with a ResNet-18 encoder achieves 99.44% and 94.35% respectively, despite being trained only on the primitive operations $\pm$1. Our results suggest that world models can learn algebraic rules when the latent transformations encode the underlying modular structure. Code is available \href{https://github.com/DL-World-Models/brojepa}{here}.
Human-Computer Interaction and Visualization
The visionary PhysioNet platform launched 25 years ago, based on a system developed at MIT in the 1970s. It has become one of the most comprehensive biomedical and clinical data repositories in existence.
Machine Intelligence
Known for his clear and elegant writing style, Bertsekas shaped fields from control and optimization to large-scale computation and artificial intelligence.
PhD student Rachel Sava, winner of the Envisioning the Future of Computing Prize, explores transformative improvements and dystopian risks of neural technology.
Computer scientist Phillip Isola cuts through the hype to explain how AI agents work and what the future might hold for this rapidly advancing technology.
