Weight Oracles: Reading Neural Network Weights with Language Models

cs.LG, cs.CR

Submitted: 2026-10-05

Updated: 2026-10-05

Comments: Spotlight at the NeurIPS 2026 Workshop on Neural Network Artifacts as a New Data Modality (NeuralArtifacts), Paris. 14 pages, 10 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• LoRA as Oracle
• Unelicitable Backdoors in Language Models via Cryptographic Transformer Circuits
• Learning to Interpret Weight Differences in Language Models
• BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
• Solving Trojan Detection Competitions with Linear Weight Classification
• Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
• Progress measures for grokking via mechanistic interpretability
• LatentQA: Teaching LLMs to Decode Activations Into Natural Language
• Z-PEFT: Zero-shot Backdoor Detection in Parameter-Efficient Fine-Tuning via Canonical Spectral Signatures
• Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets
• Weight space Detection of Backdoors in LoRA Adapters
• Position: Weight Space Should Be a First-Class Generative AI Modality

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks