Tag
This paper introduces a geometric fingerprinting framework to trace LLM lineage by analyzing spectral energy and subspace alignment in weight space, enabling discrimination between independent, same-series, and shared-base models without input data.
This paper introduces a spectral shape-based metric using Heavy-Tailed Self-Regularization theory to characterize, compare, and manage large language models. The approach is data-free, computationally efficient, and scale-invariant, enabling model lineage tracing, unsupervised clustering, and performance quantification across diverse model collections.
This paper investigates whether different offline reinforcement learning losses (RFT, RIFT, DFT, Offline GRPO, DPO) for reasoning distillation produce mechanistically distinct weight updates in a small language model. Using identical math rollouts and a controlled setup with Qwen3-4B and attention-only LoRA, they find that SFT, RFT, and RIFT yield nearly colinear weight deltas, while DPO sits in a near-orthogonal subspace and achieves the highest accuracy.
Introduces WIZARD, a weight-space meta-learning framework that generates task-specific LoRA parameters for frozen VLA policies from language instructions and demonstration videos, enabling efficient task adaptation without fine-tuning.
This paper introduces the concept of Access Sets to budget expert reads, enabling scalable weight-space model merging.