Measuring and Reducing Cross-Vendor Mismatch in Language Models
cs.LG, cs.DC
Submitted: 2026-10-04
Updated: 2026-10-04
Code: https://github.com/crova-project/crova
Sources
- Log Probability Tracking of LLM APIs
- MindVL: Towards Efficient and Effective Training of Multimodal Large Language Models on Ascend NPUs
- MarginGate: Sparse Margin-Triggered Verification for Batch-Invariant LLM Inference
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- Accelerating the Mitigation of LLM Inference Nondeterminism Across GPU Architectures
- LLM-42: Enabling Determinism in LLM Inference with Verified Speculation
- Distilling the Knowledge in a Neural Network
- Stabilizing MoE Reinforcement Learning by Aligning Training and Inference Routers
- Defeating the Training-Inference Mismatch via FP16
- The Llama 3 Herd of Models
- Beyond Precision: Training-Inference Mismatch is an Optimization Problem and Simple LR Scheduling Fixes It
- Deterministic Inference across Tensor Parallel Sizes That Eliminates Training-Inference Mismatch
- Diagnosing Training Inference Mismatch in LLM Reinforcement Learning via a Zero-Mismatch Reference
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