Image de couverture de livre

Reasoning Stabilization Point: A Training-Time Signal for Stable Evidence and Shortcut Reliance

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteur principal:	Dhayalkar, Sahil Rajesh
Format:	Preprint
Publié:	2026
Sujets:	Artificial Intelligence Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Attention as Binding: A Vector-Symbolic Perspective on Transformer Reasoning
par: Dhayalkar, Sahil Rajesh
Publié: (2025)

A Combinatorial Theory of Dropout: Subnetworks, Graph Geometry, and Generalization
par: Dhayalkar, Sahil Rajesh
Publié: (2025)

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

The Geometry of ReLU Networks through the ReLU Transition Graph
par: Dhayalkar, Sahil Rajesh
Publié: (2025)

Neural Networks as Universal Finite-State Machines: A Constructive Deterministic Finite Automaton Theory
par: Dhayalkar, Sahil Rajesh
Publié: (2025)

Logic-Gated Time-Shared Feedforward Networks for Alternating Finite Automata: Exact Simulation and Learnability
par: Dhayalkar, Sahil Rajesh
Publié: (2026)

A Constructive Framework for Nondeterministic Automata via Time-Shared, Depth-Unrolled Feedforward Networks
par: Dhayalkar, Sahil Rajesh
Publié: (2025)

Symbolic Feedforward Networks for Probabilistic Finite Automata: Exact Simulation and Learnability
par: Dhayalkar, Sahil Rajesh
Publié: (2025)

Discrete Functional Geometry of ReLU Networks via ReLU Transition Graphs
par: Dhayalkar, Sahil Rajesh
Publié: (2025)

Human Activity Recognition using Smartphones
par: Sonawane, Mayur, et autres
Publié: (2024)

Tula: Optimizing Time, Cost, and Generalization in Distributed Large-Batch Training
par: Tyagi, Sahil, et autres
Publié: (2026)

A Neuro-Symbolic Benchmark Suite for Concept Quality and Reasoning Shortcuts
par: Bortolotti, Samuele, et autres
Publié: (2024)

BEARS Make Neuro-Symbolic Models Aware of their Reasoning Shortcuts
par: Marconato, Emanuele, et autres
Publié: (2024)

Symbol Grounding in Neuro-Symbolic AI: A Gentle Introduction to Reasoning Shortcuts
par: Marconato, Emanuele, et autres
Publié: (2025)

The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMs
par: Chen, Zihan, et autres
Publié: (2026)

Gradient-based Model Shortcut Detection for Time Series Classification
par: Ibarra, Salomon, et autres
Publié: (2025)

RACE Attention: A Strictly Linear-Time Attention Layer for Training on Outrageously Large Contexts
par: Joshi, Sahil, et autres
Publié: (2025)

A Decomposable Forward Process in Diffusion Models for Time-Series Forecasting
par: Caldas, Francisco, et autres
Publié: (2026)

Reasoning-Aware Training for Time Series Forecasting
par: Ahamed, Md Atik, et autres
Publié: (2026)

Stable Reasoning, Unstable Responses: Mitigating LLM Deception via Stability Asymmetry
par: Zhang, Guoxi, et autres
Publié: (2026)

FiLoRA: Focus-and-Ignore LoRA for Controllable Feature Reliance
par: Chung, Hyunsuk, et autres
Publié: (2026)

Regulating Model Reliance on Non-Robust Features by Smoothing Input Marginal Density
par: Yang, Peiyu, et autres
Publié: (2024)

Shortcut to Nowhere: Demystifying Deep Spurious Regression
par: Xu, Guanrong, et autres
Publié: (2026)

Mitigating Shortcut Learning with InterpoLated Learning
par: Korakakis, Michalis, et autres
Publié: (2025)

Learning with Logical Constraints but without Shortcut Satisfaction
par: Li, Zenan, et autres
Publié: (2024)

TimeMaster: Training Time-Series Multimodal LLMs to Reason via Reinforcement Learning
par: Zhang, Junru, et autres
Publié: (2025)

From What to How: Attributing CLIP's Latent Components Reveals Unexpected Semantic Reliance
par: Dreyer, Maximilian, et autres
Publié: (2025)

Rectifying Shortcut Behaviors in Preference-based Reward Learning
par: Ye, Wenqian, et autres
Publié: (2025)

Towards Faithful Explanations: Boosting Rationalization with Shortcuts Discovery
par: Yue, Linan, et autres
Publié: (2024)

Building Expressive and Tractable Probabilistic Generative Models: A Review
par: Sidheekh, Sahil, et autres
Publié: (2024)

Adaptive Overclocking: Dynamic Control of Thinking Path Length via Real-Time Reasoning Signals
par: Jiang, Shuhao, et autres
Publié: (2025)

Fairness is Not Flat: Geometric Phase Transitions Against Shortcut Learning
par: Rodriguez-Alvarez, Nicolas, et autres
Publié: (2026)

Inductive Subgraphs as Shortcuts: Causal Disentanglement for Heterophilic Graph Learning
par: Wang, Xiangmeng, et autres
Publié: (2026)

MiMu: Mitigating Multiple Shortcut Learning Behavior of Transformers
par: Zhao, Lili, et autres
Publié: (2025)

Stabilizing Recurrent Dynamics for Test-Time Scalable Latent Reasoning in Looped Language Models
par: Yang, Xiao-Wen, et autres
Publié: (2026)

Beyond Low-rank Decomposition: A Shortcut Approach for Efficient On-Device Learning
par: Nguyen, Le-Trung, et autres
Publié: (2025)

EntroPIC: Towards Stable Long-Term Training of LLMs via Entropy Stabilization with Proportional-Integral Control
par: Yang, Kai, et autres
Publié: (2025)

CURE: Controlled Unlearning for Robust Embeddings -- Mitigating Conceptual Shortcuts in Pre-Trained Language Models
par: Kocak, Aysenur, et autres
Publié: (2025)

Reasoning Dynamics and the Limits of Monitoring Modality Reliance in Vision-Language Models
par: Villegas, Danae Sánchez, et autres
Publié: (2026)

RationalRewards: Reasoning Rewards Scale Visual Generation Both Training and Test Time
par: Wang, Haozhe, et autres
Publié: (2026)