{"paper":{"title":"VIP-COP: Context Optimization for Tabular Foundation Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"VIP-COP estimates importance of training samples and features to build better contexts for tabular foundation models at test time.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Leman Akoglu, Xueying Ding, Yilong Chen","submitted_at":"2026-05-13T02:28:31Z","abstract_excerpt":"Tabular foundation models (TFMs) have emerged as a powerful paradigm for in-context learning on structured data, enabling direct prediction on new tabular tasks without task-specific training. However, their effectiveness is constrained by context length limits, restricting application to medium-scale data and degrading performance when inference-time data exceed pretraining size distributions. Our work introduces VIP-COP, estimating the Value of Importance for Prediction of training examples and features for hard Context OPtimization for TFMs. Its explicit selection mechanism suppresses noise"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"VIP-COP consistently outperforms heuristic and optimized baselines across large-scale high-dimensional testbeds, including data augmentation and data-noise settings, establishing a new state of the art in test-time context refinement for TFMs.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the online KernelSHAP-based regression accurately identifies influential samples and features for prediction even when the model is treated as a black box and when data distributions differ from pretraining.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"VIP-COP is a black-box method that optimizes context for tabular foundation models by ranking and selecting high-value samples and features via online KernelSHAP regression, outperforming baselines on large high-dimensional data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"VIP-COP estimates importance of training samples and features to build better contexts for tabular foundation models at test time.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"94db3dc9acdb28ac3cfc64ff7b3e44307d1e667aa9fdb7d0615e83339180b4cc"},"source":{"id":"2605.12904","kind":"arxiv","version":1},"verdict":{"id":"4674f5b8-ebf0-4490-a651-fa06a4b68fba","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:04:24.174401Z","strongest_claim":"VIP-COP consistently outperforms heuristic and optimized baselines across large-scale high-dimensional testbeds, including data augmentation and data-noise settings, establishing a new state of the art in test-time context refinement for TFMs.","one_line_summary":"VIP-COP is a black-box method that optimizes context for tabular foundation models by ranking and selecting high-value samples and features via online KernelSHAP regression, outperforming baselines on large high-dimensional data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the online KernelSHAP-based regression accurately identifies influential samples and features for prediction even when the model is treated as a black box and when data distributions differ from pretraining.","pith_extraction_headline":"VIP-COP estimates importance of training samples and features to build better contexts for tabular foundation models at test time."},"references":{"count":42,"sample":[{"doi":"","year":2017,"title":"MIT press Cambridge, MA, USA, 2017","work_id":"727282a3-5f5f-49da-900b-bdc191bb2216","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Unleashing the potential of prompt engineering for large language models.Patterns, 6(6)","work_id":"df3167d9-81bd-4bf3-83cb-c97b9de6eeca","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Extending Context Window of Large Language Models via Positional Interpolation","work_id":"c8b6df85-e7da-4bd8-90a4-d309cc2a0f60","ref_index":3,"cited_arxiv_id":"2306.15595","is_internal_anchor":true},{"doi":"","year":2023,"title":"Longlora: Efficient fine-tuning of long-context large language models","work_id":"5649158b-b4f0-42fc-a7ce-1e8e12f1ff6d","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"Cubuk, Barret Zoph, Dandelion Mané, Vijay Vasudevan, and Quoc V","work_id":"360d57be-26bf-498a-b081-e9c46e0df1ff","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":42,"snapshot_sha256":"78df67abba004fc5716376660e5bb5b615e65d8bf77ed2210724e364e495a034","internal_anchors":6},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2e06fdef24a2073c96fffc3ff84a72db2764e980159be9d9e23c5634120ef22d"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}