arXiv subjects

Aryan Chowdhury

Publications and source records attributed to Aryan Chowdhury.

2 recordsLinked to original sources

arXiv2026

Fewer Heads, Fewer Memories: Pruning Redundant Attention Heads Suppresses Verbatim Memorisation

Large language models have been shown to memorise and reproduce verbatim sequences from their training data, raising concerns about privacy leakage, intellectual property exposure, and the limits of generalisation. Recent theoretical work suggests that attention heads may contribute to this behaviour by functioning as associative memory mechanisms, with memorisation capacity scaling with the number of heads. At the same time, empirical work has shown that many attention heads are redundant and can be pruned with limited impact on downstream performance. Despite these parallel findings, the effect of pruning redundant heads on memorised-sequence reproduction remains unexplored. In this work, we demonstrate that zeroing out a substantial fraction of attention heads prior to inference can significantly suppress verbatim memorisation while preserving, or even improving, language modelling performance on the benchmark used to identify redundant heads. Evaluating GPT-2-XL(1.5B) on known memorised sequences, we find that pruning 729 task-redundant heads reduces the memorisation count by 91.5%, while improving WikiText-103 perplexity by 0.92. These findings suggest that attention-head redundancy can be exploited to mitigate memorisation at inference time, without retraining or modifying the training data.

cs.CL↗

arXiv2026

MOSAIC: Addressing Disproportionate Task-Modality Dependencies via Modality-Partitioned Gradient Routing

In multi-task settings, combining multiple modalities allows task representations to draw on signals that a single modality alone would not capture. However, individual tasks may depend disproportionately on particular modalities, a structure that conventional fusion-based architectures do not model explicitly, instead enforcing a unified representation space. Consequently, a task's gradients can affect parameters tied to modalities it barely relies on, leaving features that compromise between competing demands. Although prior work has explored task-weighting and loss-balancing strategies, their effectiveness is limited by the absence of explicit modality separation, preventing models from optimising each modality according to its task-specific importance. To address these limitations, we propose MOSAIC (MOdality-Separated Architecture with Inter-modal Cross-attention), which preserves unimodal representations while enabling controlled interaction between them, and uses modality-specific losses to route gradients to their corresponding partitions. On the Corpus of Everyday Japanese Conversation (CEJC), MOSAIC improves average macro-F1 by 7.4 points across four tasks with a 72% reduction in backbone parameters relative to a fusion-based baseline. Our results demonstrate that modality partitioning yields consistent gains across tasks while requiring far fewer backbone parameters than a fusion-based design.

cs.LG↗