Skip to main content
Search

Hebrew University of Jerusalem

8 related articles

Subscribe to updates
  1. arXiv

    Gaussian splats skip the rasterizer: GaussianRenderFormer renders directly, reaching 38.08 dB on unseen objects

    GaussianRenderFormer replaces the hand-written rasterizer in 3D Gaussian splatting with a transformer: one token per Gaussian, and a view-dependent decoder that turns small patches of camera rays into pixels. Trained on 26,820 Objaverse objects, it reaches 38.08 dB PSNR on 300 unseen objects against renders of the full splat, while rasterizing the same pruned input reaches 44.02 dB; diagnostics point to a view-decoder bottleneck, though under-training is not ruled out.

  2. bioRxiv

    F-box proteins lose their rigid-scaffold image: cryo-EM, cross-linking and deep learning show FBXL2's substrate domain swinging 43 Å

    F-box proteins are the substrate-recruiting modules of SCF ubiquitin ligases and have long been treated as relatively rigid structural adaptors. This work combines cryo-EM structures, cross-linking mass spectrometry distance restraints and deep-learning conformational ensemble prediction to compare the same proteins across different complexes. FBXL2's substrate-binding domain moves 43 Å relative to its anchor domain, FBXL17 differs by 17 to 33 Å across assembly contexts, and FBXL1 shifts only 4.5 Å; FBXL2 needs at least four states to satisfy its cross-links to CSN.

  3. bioRxiv

    Single-cell perturbation prediction stops chasing the average: scDisentangle keeps each cell's identity in its counterfactuals

    Single-cell perturbation-response prediction has struggled to beat simple mean baselines, and population-average metrics cannot show whether a prediction keeps the source cell's identity. scDisentangle splits each cell's expression into attribute embeddings and a sequence of residual factors extracted one after another, then generates counterfactuals at inference by replacing only the attribute embeddings while holding cell-specific factors fixed. On interferon stimulation, liver infection and unseen combinatorial genetic perturbations it beat the models and mean baselines compared in each benchmark, and in a myocarditis cohort it predicted donor-specific heart-infiltrating CD8+ T cell states from peripheral blood.

  4. arXiv

    Babylonian Astronomical Diaries no longer dated tablet by tablet: a language model reads the broken text and astronomical tools rank 701 candidate years

    The Babylonian Astronomical Diaries record the sky for nearly six centuries, and many broken tablets have lost their date formula, which specialists previously dated one tablet at a time by comparing the surviving observations with computed phenomena. The new system turns observations into fixed-field clauses with a rule-based parser, has a language model check them against the text and add missed ones, and tests the clauses in all 701 years from 626 BCE to AD 75. On 120 dated tablets with decisive clauses, the top-ranked group contains the published year in 96% of cases; for 86 of them the group is a single year, the published year in 84.

  5. aclanthology.org

    Geometry proofs stop relying on the model alone: retrieve structurally similar solved problems, then let a symbolic verifier correct each step

    When a general-purpose model writes a geometry proof, it first retrieves structurally similar solved problems as examples, and a symbolic verifier then points out errors round by round. On 50 SAT-level Euclidean geometry problems from FormalGeo-7k, the full pipeline reaches 68%–96% proof accuracy, against 10%–44% for baselines that use neither analogical retrieval nor verifier feedback. Under the same verifier feedback and inference budget, analogical guidance alone lifts GPT-5 from 88% to 96% and o1 from 52% to 80%.

  6. arXiv

    OSFP4 jointly optimizes diagonal smoothing and block scales, reaching 77.03 average W4A4 accuracy on Llama-3.1-8B-Instruct while retaining about 94-97% of vendor NVFP4 prefill throughput

    The work introduces OSFP4: by analyzing a multiplicatively dithered FP4 quantizer it obtains a differentiable approximation of the matrix-product quantization error, then jointly optimizes a diagonal smoothing matrix and weight block scales per linear projection, with separate losses for RTN and GPTQ-style SIC rounding; on Llama-3.1-8B-Instruct the W4A4 configuration attains the highest average accuracy among the evaluated methods while retaining roughly 94-97% of vendor NVFP4 prefill throughput.

  7. bioRxiv

    Large Language Models Predict Human Social Behavior via Interpretable Mechanisms

    This study introduces MindEvolve, an autonomous workflow in which multiple large language models generate interpretable symbolic models of cognition for a battery of socioeconomic games spanning four core domains of social cognition—economic preferences, social preferences, social reasoning (theory of mind), and recursive planning—with expert human evaluators assessing interpretability and theoretical coherence; results show that most LLMs robustly capture economic and social preferences in relatively simple strategic settings and in some cases generate novel models integrating broader knowledge than those proposed by human experts, while their capacity to model more complex psychological processes remains limited, though a subset of state-of-the-art models shows promising performance on h

  8. medRxiv

    AI-Based Synthetic Data in Biomedicine: A Decade of Growth and a Persistent Translation Gap

    This study conducted a systematic mapping and bibliometric analysis of 4,143 publications from 2015 to 2025 on AI-generated synthetic data in biomedicine, combining expert annotation with LLM-assisted classification across data modality, medical domain, paper type, deployment status, and research stance, finding continuous growth in publication volume, 77.8% of papers strongly supportive with critical work below 1%, medical imaging dominating the corpus, highly cited primary research concentrated in molecular and pharmaceutical applications, and only 27 publications reporting operational use, thereby revealing a gap between methodological growth and deployment.

You’re all caught up