From Sycophancy (not) to Sandbagging
Felix Hofstätter, Daniel Tan, Sohaib Imran, David Quarel · Team The Sandbaggers
Submitted to Deception Detection Hackathon: Preventing AI deception. Sprint projects are early-stage work by participants, not Apart Research publications.
We investigate zero-shot generalization of sycophancy to sandbagging. We develop an evaluation suite and test harness for Huggingface models, consisting of a simple sycophancy evaluation dataset and a more advanced sandbagging evaluation dataset. For Llama3-8b-Instruct, we do not find evidence that reinforcing sycophantic behavior in the first environment generalizes to an increase in zero-shot sandbagging in the second environment.
Reviews
No public critique yet.
Cite this project
@misc{hofstatter2024from,
title = {{From Sycophancy (not) to Sandbagging}},
author = {Felix Hofstätter and Daniel Tan and Sohaib Imran and David Quarel},
year = {2024},
month = jun,
note = {Submitted to Deception Detection Hackathon: Preventing AI deception, an Apart Research Sprint},
howpublished = {\url{https://apartresearch.com/sprints/projects/from-sycophancy-(not)-to-sandbagging}},
url = {https://apartresearch.com/sprints/projects/from-sycophancy-(not)-to-sandbagging}
}More from Deception Detection Hackathon: Preventing AI deception
- 1st place by peer reviewView project: Sandbag Detection through Model Degradation
Sandbag Detection through Model Degradation
Team Truth Serum
We propose a novel technique to detect sandbagging in LLMs by adding varying amount of noise to model weights and monitoring performance.
- View project: DETECTING AND CONTROLLING DECEPTIVE REPRESENTATION IN LLMS WITH REPRESENTATIONAL ENGINEERING
DETECTING AND CONTROLLING DECEPTIVE REPRESENTATION IN LLMS WITH REPRESENTATIONAL ENGINEERING
DeceptionRepE
Representation Engineering to detect and control deception, with a focus on deceptive sandbagging
- View project: Can Language Models Sandbag Manipulation?
Can Language Models Sandbag Manipulation?
Can Language Models Sandbag Manipulation?
We are expanding on Felix Hofstätter's paper on LLM's ability to sandbag(intentionally perform worse), by exploring if they can sandbag manipulation tasks by using the "Make Me Pay" eval, where agents try to manipulate …