Leaderboards, eval harnesses and ablations — the contested business of deciding which model is actually better.
Vessel perception from space is crucial for a wide range of maritime applications, from traffic monitoring to environmental protection. However, most existing datasets predominantly focus on general o...
High-resolution visual question answering often fails because a multimodal model does not acquire the small, spatially localized evidence needed to answer a question. Sequential zooming can recover de...
Simulation-based inference is challenging when many heterogeneous observations must be composed, hierarchical latent structure must be preserved, and the simulator is misspecified relative to observed...
Recent advances in generative planning have made trajectory inpainting a promising approach to offline goal-conditioned reinforcement learning. However, these methods typically specify the planning ho...
As a fundamental image processing task, object Re-Identification (ReID) aims to retrieve objects across non-overlapping cameras. Recently, with the development of deep learning, significant advancemen...
Diffusion-based video restoration recovers realistic details, but its practical deployment is limited by two efficiency bottlenecks: costly VAE encoding and decoding, and the quadratic cost of full se...
One-dimensional (1D) variable-length visual tokenizers enable adaptive compression by varying the number of tokens, allowing downstream autoregressive (AR) models to flexibly trade off generation qual...
Multi-agent LLM systems are increasingly structured like organizations, with roles, critics and shared memory, yet they are evaluated on tasks that last minutes. We ask how such an organization behave...
Recursive self-improvement (RSI) aims to achieve compounding gains by having models improve themselves. While most existing RSI systems optimize external agent harnesses or prompts around a frozen bas...
Efficient exploration in high-dimensional decision spaces remains a central challenge for decision-making systems. Humans, in contrast, can navigate large decision spaces with remarkable efficiency. R...
The paper introduces robust successor features, a method that extends the successor representation to handle uncertainty in both reward functions and transition kernels within linear Markov Decision Processes. It provides a theoretical bound on Generalized Policy Improvement that quantifies performance loss due to mismatched dynamics, and demonstrates the approach on grid-based benchmarks against prior methods that consider only reward or transition differences.
The paper explores how data from fixed A/B tests can guide the deployment of adaptive experiments using contextual bandits. By combining off‑policy evaluation with a controlled warm‑start simulation, the authors rank pre‑specified adaptive and non‑adaptive policies using doubly robust estimators. Experiments on synthetic trials and real benchmarks show that adaptive, context‑aware policies outperform fixed allocations when heterogeneity exists, but offer little advantage otherwise.
The paper presents MoCoP v2, an enhanced contrastive pretraining method that aligns small molecule embeddings with deep‑learning‑derived cell morphology profiles. By replacing CellProfiler fingerprints with richer image‑encoded features, the new embeddings better capture how molecules alter cell morphology, leading to improved QSAR, toxicity, ADME, and activity predictions. Performance scales log‑linearly with training data size, indicating further gains with larger datasets.
GyroNovo is a new framework for de novo peptide sequencing that improves fragment imputation by guiding the process with decoder errors observed during training. It introduces mass-aware attention using rotary embeddings to encode pairwise mass differences between spectral peaks, and creates easy and hard augmented views of spectra to train the decoder under varying corruption levels. Experiments on NovoBench demonstrate significant gains, with about 9 percentage points higher peptide-level precision and 7 percentage points higher amino-acid-level precision compared to the state-of-the-art baseline.
PixSim is an open‑source simulator that models Brazil’s Pix instant‑payment system, incorporating irreversible settlement, a regulated recovery mechanism (MED), downstream fund dispersal, and capacity‑constrained analyst review. Calibrated to Banco Central do Brasil data, the model reproduces the 2025 recovery rate within 0.006 and its decomposition within 0.02, and projects future performance under various policy scenarios. Benchmark tests show how staffing levels, alert expiry rules, and scorer quality affect recovery outcomes and false‑block harm.
The paper introduces PAM-GS, a physics-aware gradient surgery technique for Physics-Informed Neural Networks (PINNs). It addresses the highly imbalanced multi-task optimisation problem in PINNs by adaptively mitigating task interference based on observed gradient conflicts. Experiments on four PDE benchmarks show that PAM-GS achieves competitive solution accuracy while maintaining strong task-balanced performance, outperforming existing methods on most problems.
The paper presents a hierarchical causal representation learning framework that models both internal climate variability and forced responses in sea surface temperature fields from a global climate model. By training on future climate change scenarios, the method accurately predicts long‑term global mean and regional temperature evolution and reproduces realistic responses to perturbations in greenhouse gas and aerosol concentrations on unseen scenarios. This demonstrates the potential of causal representation learning to improve climate model emulation.
ALF is a modular active learning framework designed to streamline the entire data acquisition process for scientific discovery. It offers a single API that supports both offline benchmarking against existing datasets and online deployment with an oracle for real‑world candidate acquisition. The framework is open‑source and available on GitHub.
The paper introduces READ, a method for composing low‑rank adapters (LoRA) in large language models. By rewriting each adapter into a balanced canonical form and enforcing a one‑directional coupling, READ allows new skills to read but never write into the output subspaces of existing skills, eliminating interference. Experiments on four benchmark suites and two model families show that READ consistently outperforms existing baselines, improving SuperGLUE scores by over twenty points and domain suite scores by more than seven points.
The paper introduces a unified pipeline that classifies univariate time series by first converting them into graphs using one of five constructions from three families (visibility, transition, proximity). The resulting graph is turned into a dissimilarity matrix, from which a Vietoris–Rips filtration produces persistence diagrams that are vectorized via persistence landscapes and topological summary statistics. Experiments on twelve UCR benchmarks reveal that no single graph construction dominates, diffusion distance consistently outperforms shortest-path metrics, and persistence-based features remain robust to noise.