Image de couverture de livre

Dynamic Proxy-Mixing: Transferring Replay Controllers from Small to Large Models for Continual Instruction Tuning

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Shihab, Ibne Farabi, Afrin, Fariya, Sharma, Anuj
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

EST-PRM: Stress-Testing Process Reward Models Before They Become Load-Bearing
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Continual Calibration: Coverage Can Collapse Before Accuracy in Lifelong LLM Fine-Tuning
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Detecting Proxy Gaming in RL and LLM Alignment via Evaluator Stress Tests
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

Calibrated Adaptation: Bayesian Stiefel Manifold Priors for Reliable Parameter-Efficient Fine-Tuning
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Cache-Efficient Posterior Sampling for Reinforcement Learning with LLM-Derived Priors Across Discrete and Continuous Domains
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

Grassmannian Mixture-of-Experts: Concentration-Controlled Routing on Subspace Manifolds
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Canonicalized Stable-List Replay for Private Federated Continual Learning over Language-Model Embeddings
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Beyond Accuracy: A Unified Random Matrix Theory Diagnostic Framework for Crash Classification Models
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Why Colors Make Clustering Harder:Global Integrality Gaps, the Price of Fairness, and Color-Coupled Algorithms in Chromatic Correlation Clustering
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Topology-Aware State Abstraction with Tangle Cores for Markov Decision Processes
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Coverage-Based Calibration for Post-Training Quantization via Weighted Set Cover over Outlier Channels
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Learning to Forget Attention: Memory Consolidation for Adaptive Compute Reduction
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Auditing Near-Optimal Policies Can Be Exponentially Hard: Conditional Query Lower Bounds via Occupancy Rashomon Capacity
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Selective Risk Certification for LLM Outputs via Information-Lift Statistics: PAC-Bayes, Robustness, and Skeleton Design
par: Akter, Sanjeda, et autres
Publié: (2025)

Efficient Unstructured Pruning of Mamba State-Space Models for Resource-Constrained Environments
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

Minimax Optimality and Spectral Routing for Majority-Vote Ensembles under Markov Dependence
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Certificate-Guided Pruning for Stochastic Lipschitz Optimization
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Beyond Variance: Knowledge-Aware LLM Compression via Fisher-Aligned Subspace Diagnostics
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

CalPro: Prior-Aware Evidential--Conformal Prediction with Structure-Aware Guarantees for Protein Structures
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

On the Sparsifiability of Correlation Clustering: Approximation Guarantees under Edge Sampling
par: Shihab, Ibne Farabi, et autres
Publié: (2026)

Anytime-Valid Answer Sufficiency Certificates for LLM Generation via Sequential Information Lift
par: Akter, Sanjeda, et autres
Publié: (2025)

What Fundamental Structure in Reward Functions Enables Efficient Sparse-Reward Learning?
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

HMAE: Self-Supervised Few-Shot Learning for Quantum Spin Systems
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

Differentiable Entropy Regularization: A Complexity-Aware Approach for Neural Optimization
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

ClearVision: Leveraging CycleGAN and SigLIP-2 for Robust All-Weather Classification in Traffic Camera Imagery
par: Sivaraman, Anush Lakshman, et autres
Publié: (2025)

Locality, Not Spectral Mixing, Governs Direct Propagation in Distributed Offline Dynamic Programming
par: Shihab, Ibne Farabi
Publié: (2026)

Causal Consistency Regularization: Training Verifiably Sensitive Reasoning in Large Language Models
par: Akter, Sanjeda, et autres
Publié: (2025)

Image Segmentation with Large Language Models: A Survey with Perspectives for Intelligent Transportation Systems
par: Akter, Sanjeda, et autres
Publié: (2025)

Large Language Models for Crash Detection in Video: A Survey of Methods, Datasets, and Challenges
par: Akter, Sanjeda, et autres
Publié: (2025)

Adaptive Constraint Propagation: Scaling Structured Inference for Large Language Models via Meta-Reinforcement Learning
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

Temporal Zoom Networks: Distance Regression and Continuous Depth for Efficient Action Localization
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

Unlocking Insights Addressing Alcohol Inference Mismatch through Database-Narrative Alignment
par: Bhagat, Sudesh, et autres
Publié: (2025)

Enhancing Traffic Incident Response through Sub-Second Temporal Localization with HybridMamba
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

LLM-Guided Probabilistic Fusion for Label-Efficient Document Layout Analysis
par: Shihab, Ibne Farabi, et autres
Publié: (2025)

Accuracy is Not Agreement: Expert-Aligned Evaluation of Crash Narrative Classification Models
par: Bhagat, Sudesh Ramesh, et autres
Publié: (2025)

Heart rate and respiratory rate prediction from noisy real-world smartphone based on Deep Learning methods
par: Shihab, Ibne Farabi
Publié: (2025)

Precise and Robust Sidewalk Detection: Leveraging Ensemble Learning to Surpass LLM Limitations in Urban Environments
par: Shihab, Ibne Farabi, et autres
Publié: (2024)

MAGNET-KG: Maximum-Entropy Geometric Networks for Temporal Knowledge Graphs: Theoretical Foundations and Mathematical Framework
par: Shihab, Ibne Farabi
Publié: (2025)

DeepLocalization: Using change point detection for Temporal Action Localization
par: Rahman, Mohammed Shaiqur, et autres
Publié: (2024)

HPCR: Holistic Proxy-based Contrastive Replay for Online Continual Learning
par: Lin, Huiwei, et autres
Publié: (2023)