5 citations · 6 across the 3 of their papers we have counts for
5 papers
A Dark Forest First Attack Is Rational Only If the Attacker Accepts It as Its Last
Harvey Dam
The Dark Forest argument holds that a civilization that detects another should strike it at once, so detection, and any transmission that invites it, is danger. Existing models mak…
Derailing Non-Answers via Logit Suppression at Output Subspace Boundaries in RLHF-Aligned Language Models
Harvey Dam, Jonas Knochelmann, Vinu Joseph +1
We introduce a method to reduce refusal rates of large language models (LLMs) on sensitive content without modifying model weights or prompts. Motivated by the observation that ref…
Local-Order Auxiliary Losses Can Improve Autoencoder Reconstruction
Harvey Dam, Martin Burtscher, Tripti Agarwal +1
Mean-squared error is the default objective for training autoencoders, yet compressed reconstructions often depend not only on pointwise accuracy but also on preserving local spati…
What Operations can be Performed Directly on Compressed Arrays, and with What Error?
Tripti Agarwal, Harvey Dam, Dorra Ben Khalifa +3
In response to the rapidly escalating costs of computing with large matrices and tensors caused by data movement, several lossy compression methods have been developed to significa…
MPGemmFI: A Fault Injection Technique for Mixed Precision GEMM in ML Applications
Bo Fang, Xinyi Li, Harvey Dam +9
Emerging deep learning workloads urgently need fast general matrix multiplication (GEMM). To meet such demand, one of the critical features of machine-learning-specific accelerator…