Barsat Khadka

Barsat Khadka

Mechanistic interpretability has identified small sets of attention heads that implement specific behaviours in transformer language models, but recovering these circuits typically requires a bespoke analytical pipeline for each new task. We recast circuit discovery as a reinforcement-learning problem. An agent operates over the 144 attention heads of GPT-2 small as a discrete action space; each action triggers a zero-ablation and a contrastive reward that subtracts the ablation's damage to general next-token prediction from its damage to the target task. A single PPO policy, trained on two tasks (induction and IOI) in a vectorised multi-task environment, attains the per-episode oracle on both training tasks and on a held-out third task (docstring completion). Its preferred heads coincide with the canonical heads of established literature on precisely the axes those papers identify as causally non-redundant under single-head ablation; the categories they identify as redundant are correctly de-prioritised by the agent. On the held-out task, best-of-five planning recovers 96\% of the oracle ceiling with no task signal supplied at evaluation. These results indicate that reinforcement learning over causal interventions is a viable, transferable substrate for identifying the single-head bottlenecks of mechanistic circuits, complementary to existing path-patching approaches.

Barsat Khadka, Prasant Koirala, Kshitiz Neupane et al.

Social engineering attacks exploit human trust rather than software vulnerabilities, making them difficult to detect using conventional filters. We propose a two-stage filter-then-verify framework combining inductive Graph Neural Networks (GNNs) for structural anomaly detection with a co-attention ModernBERT model for content verification. The GNN identifies anomalous sender-receiver patterns, while BERT analyzes message context to reduce false positives. Using the Enron dataset augmented with realistic synthetic campaigns, we show that the framework achieves 86% recall in structural filtering and over 92% precision after BERT refinement, effectively detecting both external attacks and insider threats. Our results demonstrate that combining structural and content analysis allows practical, scalable detection of multi-stage social engineering attacks in email networks.

Barsat Khadka, Kawsher Roxy, Md Rubel Ahmed

Graph Neural Networks (GNNs) are increasingly explored for physical design analysis in Electronic Design Automation, particularly for modeling Clock Tree Synthesis behavior such as clock skew and buffering complexity. However, practical deployment remains limited due to the prohibitive memory and runtime cost of operating on raw gate-level netlists. Graph coarsening is commonly used to improve scalability, yet its impact on CTS-critical learning objectives is not well characterized. This paper introduces CTS-Bench, a benchmark suite for systematically evaluating the trade-offs between graph coarsening, prediction accuracy, and computational efficiency in GNN-based CTS analysis. CTS-Bench consists of 4,860 converged physical design solutions spanning five architectures and provides paired raw gate-level and clustered graph representations derived from post-placement designs. Using clock skew prediction as a representative CTS task, we demonstrate a clear accuracy-efficiency trade-off. While graph coarsening reduces GPU memory usage by up to 17.2x and accelerates training by up to 3x, it also removes structural information essential for modeling clock distribution, frequently resulting in negative $R^2$ scores under zero-shot evaluation. Our findings indicate that generic graph clustering techniques can fundamentally compromise CTS learning objectives, even when global physical metrics remain unchanged. CTS-Bench enables principled evaluation of CTS-aware graph coarsening strategies, supports benchmarking of GNN architectures and accelerators under realistic physical design constraints, and provides a foundation for developing learning-assisted CTS analysis and optimization techniques.