Arun Jose — accepted AI-conference papers

2026

Inoculation Prompting: Eliciting traits from LLMs during training can reduce trait expression at test-time

Daniel Tan, Anders Cairns Woodruff, Niels Warncke, Arun Jose, Maxime Nicolas Riché, David Demitri Africa +1

ICLR 2026poster

Language model finetuning often results in learning undesirable traits in combination with desired ones. To address this, we propose inoculation prompting: modifying finetuning data by prepending a short system-prompt instruction that deliberately elicits the undesirable trait. At test time, we eval…

Cited by 0 Source ↗Scholar ↗

2026

Strategic Obfuscation of Deceptive Reasoning in Language Models

Arun Jose, Niels Warncke, Mia Taylor

ICLR 2026poster

Large language models can exhibit different behaviors during training versus deployment, a phenomenon known as alignment faking. While prior work has shown models can explicitly reason about faking alignment to preserve their goals, we demonstrate that Claude 3.5 Sonnet exhibits alignment faking beh…

Cited by 0 Source ↗Scholar ↗

2025

Reasoning Models Sometimes Output Illegible Chains of Thought

Arun Jose

NeurIPS 2025poster

Language models trained via outcome-based reinforcement learning (RL) to reason using chain-of-thought (CoT) have shown remarkable performance. Monitoring such a model's CoT may allow us to understand its intentions and detect potential malicious behavior. However, to be effective, this requires tha…

Cited by 0 Source ↗Scholar ↗

2025

Why Do Some Language Models Fake Alignment While Others Don't?

Abhay Sheshadri, John Hughes, Julian Michael, Alex Troy Mallen, Arun Jose, Fabien Roger

NeurIPS 2025spotlight

*Alignment Faking in Large Language Models* presented a demonstration of Claude 3 Opus and Claude 3.5 Sonnet selectively complying with a helpful-only training objective to prevent modification of their behavior outside of training. We expand this analysis to 25 models and find that only 5 (Claude 3…

Cited by 0 Source ↗Scholar ↗