Customer-obsessed science
Research areas
-
November 16, 202611 min readNew open-source tool turns existing test suites into sets of formal-verification tasks, which catch buggy solutions that tests alone would miss.
-
November 11, 20265 min read
-
-
-
Featured news
-
ECML PKDD 20262026In an e-commerce setup, product inconsistency detection flags listings where the title contradicts the bullets or attributes, for example, a 'cotton shirt' title with 'polyester' bullets. It matters for catalog quality, but large models and frontier LLM APIs are too expensive to run over a full catalog. While a 1.5B student trained with SFT on ∼101K labeled listings, the student reaches F1=0.833, a 4B model
-
2026Seller-authored product titles often contradict their own bullet points (a title reading '128GB' against a bullet claiming '512GB'), which misleads shoppers and drives returns. At catalog scale, detection is precision-binding: even a small false-positive rate flags a huge absolute volume of good listings. We deploy a two-stage cascade on one national storefront of a large e-commerce marketplace: a compact
-
2026In common-context question answering (CCQA) tasks, multiple input questions share a common context to base their answers from. However, Large Language Models typically generate each answer using an independent prompt. While existing batching and caching techniques help improve parallelism and reduce repeated computations, the separation of questions across prompts limits the achievable speedup, as modern
-
2026The LLM-as-a-judge (LLMaJ) framework has emerged as a promising solution for cheap, reproducible, reference-free Natural Language Generation (NLG) evaluation. Prior work seeks to improve LLMaJ by decomposing evaluation tasks into simpler sub-tasks. In this work, we systematically compare LLMaJ methods with and without decomposition on multiple NLG datasets. We find no evidence that LLMaJ with task decomposition
-
2026Upgrading embedding models typically requires expensive database re-indexing, as new query embeddings are incompatible with existing database embeddings. While Backward Compatible Training (BCT) mitigates this by enforcing compatibility during training, existing approaches often require updating the backbone model. This is impractical because of significant training cost, the risk of performance regression
Collaborations
View allWhether you're a faculty member or student, there are number of ways you can engage with Amazon.
View all