Cover Image

Mitigating Adaptive Attacks against Reasoning Models with Activation Consistency Training

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Shah, Avidan, Brinkmann, Jannik, Angell, Rico
Format:	Preprint
Published:	2026
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Jailbreak Transferability Emerges from Shared Representations
by: Angell, Rico, et al.
Published: (2025)

On-Policy Consistency Training Improves LLM Safety with Minimal Capability Degradation
by: Han, Andy, et al.
Published: (2026)

Polynomial Precision Dependence Solutions to Alignment Research Center Matrix Completion Problems
by: Angell, Rico
Published: (2024)

A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Task
by: Brinkmann, Jannik, et al.
Published: (2024)

Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models
by: Karvonen, Adam, et al.
Published: (2024)

Efficient Mitigation of Bus Bunching through Setter-Based Curriculum Learning
by: Shah, Avidan, et al.
Published: (2024)

Fast, Scalable, Warm-Start Semidefinite Programming with Spectral Bundling and Sketching
by: Angell, Rico, et al.
Published: (2023)

In-Context Algebra
by: Todd, Eric, et al.
Published: (2025)

Consistent Valid Physically-Realizable Adversarial Attack against Crowd-flow Prediction Models
by: Ali, Hassan, et al.
Published: (2023)

DeepShare: Sharing ReLU Across Channels and Layers for Efficient Private Inference
by: Bornfeld, Yonathan, et al.
Published: (2025)

FAROS: Robust Federated Learning with Adaptive Scaling against Backdoor Attacks
by: Hu, Chenyu, et al.
Published: (2026)

Estimating Tail Risks in Language Model Output Distributions
by: Angell, Rico, et al.
Published: (2026)

Adaptive Discretization for Consistency Models
by: Bai, Jiayu, et al.
Published: (2025)

Better Call SAUL: Fluent and Consistent Language Model Editing with Generation Regularization
by: Wang, Mingyang, et al.
Published: (2024)

Watermark Smoothing Attacks against Language Models
by: Chang, Hongyan, et al.
Published: (2024)

ARMs: Adaptive Red-Teaming Agent against Multimodal Models with Plug-and-Play Attacks
by: Chen, Zhaorun, et al.
Published: (2025)

Consistency Training Helps Stop Sycophancy and Jailbreaks
by: Irpan, Alex, et al.
Published: (2025)

Mechanisms of AI Protein Folding in ESMFold
by: Lu, Kevin, et al.
Published: (2026)

Adaptive Discounting of Training Time Attacks
by: Bector, Ridhima, et al.
Published: (2024)

ZClip: Adaptive Spike Mitigation for LLM Pre-Training
by: Kumar, Abhay, et al.
Published: (2025)

MAED: Mathematical Activation Error Detection for Mitigating Physical Fault Attacks in DNN Inference
by: Ahmadi, Kasra, et al.
Published: (2026)

Fast Adversarial Training against Sparse Attacks Requires Loss Smoothing
by: Zhong, Xuyang, et al.
Published: (2025)

Beyond Training: Optimizing Reinforcement Learning Based Job Shop Scheduling Through Adaptive Action Sampling
by: de Puiseau, Constantin Waubert, et al.
Published: (2024)

A Family of Adaptive Activation Functions for Mitigating Failure Modes in Physics-Informed Neural Networks
by: Murari, Krishna
Published: (2026)

Semantic Membership Inference Attack against Large Language Models
by: Mozaffari, Hamid, et al.
Published: (2024)

Certified Self-Consistency: Statistical Guarantees and Test-Time Training for Reliable Reasoning in LLMs
by: Cordero-Encinar, Paula, et al.
Published: (2025)

Adaptive Backdoor Attacks with Reasonable Constraints on Graph Neural Networks
by: Dong, Xuewen, et al.
Published: (2025)

Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning
by: Kim, Junseok, et al.
Published: (2026)

Self-Training Meets Consistency: Improving LLMs' Reasoning with Consistency-Driven Rationale Evaluation
by: Lee, Jaehyeok, et al.
Published: (2024)

TrojanTO: Action-Level Backdoor Attacks against Trajectory Optimization Models
by: Dai, Yang, et al.
Published: (2025)

Safe Training with Sensitive In-domain Data: Leveraging Data Fragmentation To Mitigate Linkage Attacks
by: Ignashina, Mariia, et al.
Published: (2024)

Mitigating Catastrophic Forgetting in Mathematical Reasoning Finetuning through Mixed Training
by: Reynolds, John Graham
Published: (2025)

Memory Faults in Activation-sparse Quantized Deep Neural Networks: Analysis and Mitigation using Sharpness-aware Training
by: Malhotra, Akul, et al.
Published: (2024)

PAINT: Partial-Solution Adaptive Interpolated Training for Self-Distilled Reasoners
by: Tan, Zhiquan, et al.
Published: (2026)

Defending Deep Regression Models against Backdoor Attacks
by: Du, Lingyu, et al.
Published: (2024)

Consistency Training with Physical Constraints
by: Chang, Che-Chia, et al.
Published: (2025)

Tokens for Learning, Tokens for Unlearning: Mitigating Membership Inference Attacks in Large Language Models via Dual-Purpose Training
by: Tran, Toan, et al.
Published: (2025)

The Environmental Impact of Ensemble Techniques in Recommender Systems
by: Nitschke, Jannik
Published: (2025)

FLAIN: Mitigating Backdoor Attacks in Federated Learning via Flipping Weight Updates of Low-Activation Input Neurons
by: Ding, Binbin, et al.
Published: (2024)

ActivationReasoning: Logical Reasoning in Latent Activation Spaces
by: Helff, Lukas, et al.
Published: (2025)