arXiv subjects

Harvey Dam

Publications and source records attributed to Harvey Dam.

5 recordsLinked to original sources

arXiv2025

Derailing Non-Answers via Logit Suppression at Output Subspace Boundaries in RLHF-Aligned Language Models

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 refusals in certain models were often preceded by the specific token sequence of a token marking the beginning of the chain-of-thought (CoT) block ( ) followed by a double newline token (\n\n), we investigate the impact of two simple formatting adjustments during generation: suppressing \n\n after and suppressing the end-of-sequence token after the end of the CoT block ( ). Our method requires no datasets, parameter changes, or training, relying solely on modifying token probabilities during generation. In our experiments with official DeepSeek-R1 distillations, these interventions increased the proportion of substantive answers to sensitive prompts without affecting performance on standard benchmarks. Our findings suggest that refusal behaviors can be circumvented by blocking refusal subspaces at specific points in the generation process.

cs.CL↗

arXiv2025

Local-Order Auxiliary Losses Can Improve Autoencoder Reconstruction

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 spatial order. We study whether structural auxiliary losses can improve, rather than trade off against, MSE in finite-capacity autoencoders. We introduce finite-difference sign error (FDSE), a local-order auxiliary objective that penalizes disagreements between the signs of neighboring finite differences in the target and reconstruction. FDSE is simple, architecture-agnostic, and differentiable through smooth sign surrogates. Across four tensor reconstruction tasks, we find that moderate mixtures of MSE and FDSE can substantially reduce validation MSE relative to pure MSE training. In coefficient sweeps, FDSE mixtures reduce validation MSE by 2.3$\times$--7.0$\times$ over pure MSE on these tasks, while comparisons with other auxiliary objectives show FDSE to be among the strongest structural objectives tested. The effect is not universal: pure FDSE performs poorly, and gains are largest for coherent spatial fields where local order carries information about the underlying signal. These results suggest that, in compressed-latent reconstruction, appropriately weighted local-structure supervision can guide optimization toward solutions with better pointwise accuracy, rather than merely improving perceptual or structural metrics at MSE's expense.

cs.LG↗

arXiv2024

What Operations can be Performed Directly on Compressed Arrays, and with What Error?

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 significantly reduce data volumes. Unfortunately, all these methods require the data to be decompressed before further computations are done. In this work, we develop a lossy compressor that allows a dozen fairly fundamental operations directly on compressed data while offering good compression ratios and modest errors. We implement a new compressor PyBlaz based on the familiar GPU-powered PyTorch framework, and evaluate it on three non-trivial applications, choosing different number systems for internal representation. Our results demonstrate that the compressed-domain operations achieve good scalability with problem sizes while incurring errors well within acceptable limits. To our best knowledge, this is the first such lossy compressor that supports compressed-domain operations while achieving acceptable performance as well as error.

cs.DC↗

arXiv2023

MPGemmFI: A Fault Injection Technique for Mixed Precision GEMM in ML Applications

Emerging deep learning workloads urgently need fast general matrix multiplication (GEMM). To meet such demand, one of the critical features of machine-learning-specific accelerators such as NVIDIA Tensor Cores, AMD Matrix Cores, and Google TPUs is the support of mixed-precision enabled GEMM. For DNN models, lower-precision FP data formats and computation offer acceptable correctness but significant performance, area, and memory footprint improvement. While promising, the mixed-precision computation on error resilience remains unexplored. To this end, we develop a fault injection framework that systematically injects fault into the mixed-precision computation results. We investigate how the faults affect the accuracy of machine learning applications. Based on the error resilience characteristics, we offer lightweight error detection and correction solutions that significantly improve the overall model accuracy if the models experience hardware faults. The solutions can be efficiently integrated into the accelerator's pipelines.

cs.DC↗

arXiv2023

Understanding the Effect of the Long Tail on Neural Network Compression

Network compression is now a mature sub-field of neural network research: over the last decade, significant progress has been made towards reducing the size of models and speeding up inference, while maintaining the classification accuracy. However, many works have observed that focusing on just the overall accuracy can be misguided. E.g., it has been shown that mismatches between the full and compressed models can be biased towards under-represented classes. This raises the important research question, can we achieve network compression while maintaining "semantic equivalence" with the original network? In this work, we study this question in the context of the "long tail" phenomenon in computer vision datasets observed by Feldman, et al. They argue that memorization of certain inputs (appropriately defined) is essential to achieving good generalization. As compression limits the capacity of a network (and hence also its ability to memorize), we study the question: are mismatches between the full and compressed models correlated with the memorized training data? We present positive evidence in this direction for image classification tasks, by considering different base architectures and compression schemes.

cs.LG↗