9 papers
On Mitigation of Subliminal Learning in Large Language Models
Atsushi Yanagisawa, Brendan Gho, Rajendran Ramesh Babu Manoj Narender +3
Knowledge distillation can transmit unintended behavioral traits from a teacher model to a student through training data that appear semantically unrelated to those traits, a pheno…
Probing the Origins of Reasoning Performance: Representational Quality for Mathematical Problem-Solving in RL vs. SFT Fine-Tuned Models
Antyabha Rahman, Akshaj Gurugubelli, Omar Ankit +2
Large reasoning models trained via reinforcement learning (RL) have been increasingly shown to outperform their supervised fine-tuned (SFT) counterparts on mathematical reasoning t…
Interpreting Latent CoT Reasoning as Dynamical Systems
Sabari Iyyappan Duraipandian, Shreya Sanjay Boyane, Manju Nagesh +3
Recent latent reasoning methods, such as CODI and COCONUT, face a fundamental interpretability problem: they maintain multiple superimposed candidate traces in the hidden space at…
Mitigating Forgetting in Continual Learning with Selective Gradient Projection
Anika Singh, Aayush Dhaulakhandi, Varun Chopade +3
As neural networks are increasingly deployed in dynamic environments, they face the challenge of catastrophic forgetting, the tendency to overwrite previously learned knowledge whe…
Peek-a-Boo Reasoning: Contrastive Region Masking in MLLMs
Isha Chaturvedi, Anjana Nair, Yushen Li +5
We introduce Contrastive Region Masking (CRM), a training free diagnostic that reveals how multimodal large language models (MLLMs) depend on specific visual regions at each step o…
From Competition to Coordination: Market Making as a Scalable Framework for Safe and Aligned Multi-Agent LLM Systems
Brendan Gho, Suman Muppavarapu, Afnan Shaik +6
As foundation models are increasingly deployed as interacting agents in multi-agent systems, their collective behavior raises new challenges for trustworthiness, transparency, and…
LIFT: Likelihood-Based Intervention Faithfulness Training
Anand Swaroop, Akshat Nallani, Saksham Uboweja +7
Chain-of-thought (CoT) reasoning improves LLM accuracy on complex tasks, but CoTs are often unfaithful; that is, the model does not rely on the stated reasoning steps to produce it…
Evaluation Awareness Scales Predictably in Open-Weights Large Language Models
Maheep Chaudhary, Ian Su, Nikhil Hooda +6
Large language models (LLMs) can internally distinguish between evaluation and deployment contexts, a behaviour known as \emph{evaluation awareness}. This undermines AI safety eval…
Probe-Rewrite-Evaluate: A Workflow for Reliable Benchmarks and Quantifying Evaluation Awareness
Lang Xiong, Nishant Bhargava, Jianhang Hong +4
Large Language Models (LLMs) often exhibit significant behavioral shifts when they perceive a change from a real-world deployment context to a controlled evaluation setting, a phen…