Sharadind Peddiraju

Maxime Heuillet, Sharadind Peddiraju

Data science tasks span from closed-ended information extraction to open-ended analysis, presenting significant challenges for automation. Recent AI agents powered by language models show promise for handling such complex tasks. However, existing agents typically rely on a single initial state that conditions the entire agent's execution, making them vulnerable to cascading errors initiated by a suboptimal initial state. To mitigate this, we present CIPHER, an automated data science agent that leverages test-time scaling through the generation and selection of multiple initial states for concurrent execution. Unlike existing works on test-time scaling of AI agents, CIPHER explicitly decouples the generation of candidate initial states from their strategic selection for parallel execution. Through extensive evaluation on two benchmarks (closed-form and open-form tasks), we demonstrate that CIPHER exceeds state-of-the-art performance in matched-model comparisons, and remains competitive against larger-model baselines despite relying on a substantially smaller base LM. Our empirical study characterizes the design space of the Decoupled Exploration-Selection (DES) framework: we quantify how generation strategy, selection strategy, and aggregator model capacity contribute to overall performance, and derive actionable design recommendations for practitioners.

Sharadind Peddiraju, Srini Rajagopal

The challenge of creating domain-centric embeddings arises from the abundance of unstructured data and the scarcity of domain-specific structured data. Conventional embedding techniques often rely on either modality, limiting their applicability and efficacy. This paper introduces a novel modeling approach that leverages structured data to filter noise from unstructured data, resulting in embeddings with high precision and recall for domain-specific attribute prediction. The proposed model operates within a Hybrid Collaborative Filtering (HCF) framework, where generic entity representations are fine-tuned through relevant item prediction tasks. Our experiments, focusing on the cloud computing domain, demonstrate that HCF-based embeddings outperform AutoEncoder-based embeddings (using purely unstructured data), achieving a 28% lift in precision and an 11% lift in recall for domain-specific attribute prediction.