Logits are All We Need to Adapt Closed Models
cs.LG, cs.AI, cs.CL
Submitted: 2025-02-03
Updated: 2026-09-18
Comments: Accepted to ICML 2025. 29 pages, 8 figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Constitutional AI: Harmlessness from AI Feedback
- Black-box Prompt Learning for Pre-trained Language Models
- GPT-4 Technical Report
- The Llama 3 Herd of Models
- Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
- Aligning LLM Agents by Learning Latent Preference from User Edits
- LoRA: Low-Rank Adaptation of Large Language Models
- Calibrating Long-form Generations from Large Language Models
- GPT-4o System Card
- Tuning Language Models by Proxy
- The Hessian perspective into the Nature of Convolutional Neural Networks
- On the Duality between Gradient Transformations and Adapters
- Calibrating Large Language Models Using Their Generations Only
- HuggingFace's Transformers: State-of-the-art Natural Language Processing
- To Cool or not to Cool? Temperature Network Meets Large Foundation Models via DRO
- A Study on the Calibration of In-context Learning
- Why Transformers Need Adam: A Hessian Perspective
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: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks