Imagen de Portada

Turbo Connection: Reasoning as Information Flow from Higher to Lower Layers

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Tang, Mohan, Lu, Sidi
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning Artificial Intelligence
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Can LLMs Guide Their Own Exploration? Gradient-Guided Reinforcement Learning for LLM Reasoning
por: Liang, Zhenwen, et al.
Publicado: (2025)

Reasoning LLMs are Wandering Solution Explorers
por: Lu, Jiahao, et al.
Publicado: (2025)

Dynamic Context Adaptation and Information Flow Control in Transformers: Introducing the Evaluator Adjuster Unit and Gated Residual Connections
por: Dhayalkar, Sahil Rajesh
Publicado: (2024)

TurboHopp: Accelerated Molecule Scaffold Hopping with Consistency Models
por: Yoo, Kiwoong, et al.
Publicado: (2024)

TurboAngle: Near-Lossless KV Cache Compression via Uniform Angle Quantization
por: Patel, Dipkumar
Publicado: (2026)

Bayesian Deep Learning Via Expectation Maximization and Turbo Deep Approximate Message Passing
por: Xu, Wei, et al.
Publicado: (2024)

Layer Specialization Underlying Compositional Reasoning in Transformers
por: Liu, Jing
Publicado: (2025)

Higher Gauge Flow Models
por: Strunk, Alexander, et al.
Publicado: (2025)

CHAINSFORMER: Numerical Reasoning on Knowledge Graphs from a Chain Perspective
por: Zhao, Ze, et al.
Publicado: (2025)

TurboAttention: Efficient Attention Approximation For High Throughputs LLMs
por: Kang, Hao, et al.
Publicado: (2024)

On the Limits of Layer Pruning for Generative Reasoning in Large Language Models
por: Shrestha, Safal, et al.
Publicado: (2026)

Revealing Combinatorial Reasoning of GNNs via Graph Concept Bottleneck Layer
por: Niu, Yue, et al.
Publicado: (2026)

Revisiting RaBitQ and TurboQuant: A Symmetric Comparison of Methods, Theory, and Experiments
por: Gao, Jianyang, et al.
Publicado: (2026)

Learning Tennis Strategy Through Curriculum-Based Dueling Double Deep Q-Networks
por: Mohan, Vishnu
Publicado: (2025)

Unveiling Mode Connectivity in Graph Neural Networks
por: Li, Bingheng, et al.
Publicado: (2025)

UnStar: Unlearning with Self-Taught Anti-Sample Reasoning for LLMs
por: Sinha, Yash, et al.
Publicado: (2024)

Self-Evolving Curriculum for LLM Reasoning
por: Chen, Xiaoyin, et al.
Publicado: (2025)

Beyond the Lower Bound: Bridging Regret Minimization and Best Arm Identification in Lexicographic Bandits
por: Xue, Bo, et al.
Publicado: (2025)

Between the Layers Lies the Truth: Uncertainty Estimation in LLMs Using Intra-Layer Local Information Scores
por: Badash, Zvi N., et al.
Publicado: (2026)

Disentangling Recall and Reasoning in Transformer Models through Layer-wise Attention and Activation Analysis
por: Fartale, Harshwardhan, et al.
Publicado: (2025)

Layer Importance for Mathematical Reasoning is Forged in Pre-Training and Invariant after Post-Training
por: Nepal, Aadim, et al.
Publicado: (2025)

On Variance Reduction in Learning Mean Flows
por: Lu, Juanwu, et al.
Publicado: (2026)

Efficient Knowledge Tracing Leveraging Higher-Order Information in Integrated Graphs
por: Han, Donghee, et al.
Publicado: (2025)

AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs
por: Liu, Xiaogeng, et al.
Publicado: (2024)

Efficient Regression-Based Training of Normalizing Flows for Boltzmann Generators
por: Rehman, Danyal, et al.
Publicado: (2025)

CLOVER: Cross-Layer Orthogonal Vectors Pruning and Fine-Tuning
por: Meng, Fanxu, et al.
Publicado: (2024)

Implicit Hypergraph Neural Networks: A Stable Framework for Higher-Order Relational Learning with Provable Guarantees
por: Li, Xiaoyu, et al.
Publicado: (2025)

Are Transformers Able to Reason by Connecting Separated Knowledge in Training Data?
por: Yin, Yutong, et al.
Publicado: (2025)

When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs
por: Wang, Keyu, et al.
Publicado: (2025)

Native Reasoning Models: Training Language Models to Reason on Unverifiable Data
por: Wang, Yuanfu, et al.
Publicado: (2026)

Hypergraph Pattern Machine: Compositional Tokenization for Higher-Order Interactions
por: Zhao, Kyrie, et al.
Publicado: (2026)

Higher-order Structure Boosts Link Prediction on Temporal Graphs
por: Liu, Jingzhe, et al.
Publicado: (2025)

Out-of-Distribution Adaptation in Offline RL: Counterfactual Reasoning via Causal Normalizing Flows
por: Cho, Minjae, et al.
Publicado: (2024)

Train at Moving Edge: Online-Verified Prompt Selection for Efficient RL Training of Large Reasoning Model
por: Wu, Jiahao, et al.
Publicado: (2026)

Information-Theoretic Greedy Layer-wise Training for Traffic Sign Recognition
por: Lyu, Shuyan, et al.
Publicado: (2025)

Stem: Rethinking Causal Information Flow in Sparse Attention
por: Niu, Lin, et al.
Publicado: (2026)

Permissive Information-Flow Analysis for Large Language Models
por: Siddiqui, Shoaib Ahmed, et al.
Publicado: (2024)

The Other Side of the Coin: Unveiling the Downsides of Model Aggregation in Federated Learning from a Layer-peeled Perspective
por: Zhu, Guogang, et al.
Publicado: (2025)

Lower bounds on transformers with infinite precision
por: Kozachinskiy, Alexander
Publicado: (2024)

TS-Reasoner: Domain-Oriented Time Series Inference Agents for Reasoning and Automated Analysis
por: Ye, Wen, et al.
Publicado: (2024)