برچسب: یادگیری ماشین

SIEVE: Selective Unlearning in Vision-Language Models via Attention-Value Suppression

SIEVE suppresses attention values for forget examples and matches them to a reference model for retain examples, reducing recovery of sensitive image–question–answer associations. Reported MLLMU-Bench results also show…

A Rubric Landscape for Evaluating Clinical Reasoning in Large Language Models

This preprint, whose independent peer review is unconfirmed, examines how far existing tools can evaluate clinical reasoning by large language models over longitudinal, multi-document records.

MOLE: Mixture of Latent Experts for Complementary Visual Reasoning

MOLE is designed to reduce redundancy in latent visual reasoning through selective visual routing and expert-specific latent transformations.

When Translations Need to Reach the Same Router

This research proposes aligning expert-selection distributions, rather than hidden states directly, to bring language processing closer together in MoE models—an encouraging idea with clear boundaries.

Disentangling Spatial Reasoning Failures with Composable Video Operators

This paper introduces CROSS, a training-free library for explicit geometric computations in video spatial reasoning. By separating perception, evidence sufficiency, measurement choice, and coordinate conventions, the authors report…

Cross-Lingual Alignment in MoE Models Through Router Outputs

This paper introduces a method for aligning processing across languages in decoder-only MoE models by bringing the mean expert-routing distributions of translated sequences closer together. Experiments on four…

Sphere Encoder 2: Image Generation with Rotation in a Spherical Latent Space

This paper introduces Sphere Encoder 2, a single-stage autoencoder for image generation that improves ImageNet samples by covering the equatorial region of its spherical latent space and separating…

Global Coherence in Multi-Agent Collaboration

This paper studies cases where each agent acts correctly from its local view, yet the team’s combined result violates shared constraints. Its central proposal is an external harness…

Mem++: Non-Destructive Long-Term Memory for Organizational Agents

Mem++ preserves complete organizational documents with dates and authors, deferring the choice among conflicting versions until question answering.

When the Same Reward Teaches Different Visual Skills

This preprint argues that improved visual-benchmark scores need not come from learning from images; task design must make visual evidence necessary for reward.