# ABAI at COLIEE 2026 Task 1: Multi-Stage Retrieval with GraphRAG-Enhanced Meta-Learning, and a Post-Hoc Study of the Cross-Validation-to-Test Gap

arXiv: https://arxiv.org/abs/2609.26237 · HTML: https://mlsanity.com/paper/2609.26237/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 24 references that are in the map.
- First posted on arXiv 12 Aug 2026.

## Builds on (most cited first, 12 of 12)

- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://mlsanity.com/paper/2006.03654/) (2020)
- [M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation](https://mlsanity.com/paper/2402.03216/) (2024)
- [C-Pack: Packed Resources For General Chinese Embeddings](https://mlsanity.com/paper/2309.07597/) (2023)
- [From Louvain to Leiden: guaranteeing well-connected communities](https://mlsanity.com/paper/1810.08473/) (2018)
- [Graph Retrieval-Augmented Generation: A Survey](https://mlsanity.com/paper/2408.08921/) (2024)
- [SAILER: Structure-aware Pre-trained Language Model for Legal Case Retrieval](https://mlsanity.com/paper/2304.11370/) (2023)
- [LeSICiN: A Heterogeneous Graph-based Approach for Automatic Legal Statute Identification from Indian Legal Documents](https://mlsanity.com/paper/2112.14731/) (2021)
- [Adaptive Re-Ranking with a Corpus Graph](https://mlsanity.com/paper/2208.08942/) (2022)
- [U-CREAT: Unsupervised Case Retrieval using Events extrAcTion](https://mlsanity.com/paper/2307.05260/) (2023)
- [Towards an In-Depth Comprehension of Case Relevance for Better Legal Retrieval](https://mlsanity.com/paper/2404.00947/) (2024)
- [When Should Dense Retrievers Be Updated in Evolving Corpora? Detecting Out-of-Distribution Corpora Using GradNormIR](https://mlsanity.com/paper/2506.01877/) (2025)

## Abstract (arXiv)

We present the ABAI submission to COLIEE 2026 Task 1, case law retrieval, together with a controlled study of why it underperformed. The task suppresses the cited passages themselves, which removes much of the lexical overlap a retriever would rely on. Our pipeline answers this with four independently trained stages: multi-view BM25 over citation-context windows with reciprocal rank fusion, neural reranking, graph-based features from entity communities and a graph attention network, and a LightGBM meta-learner over 34 features. Our best run reached F1=0.177 on the official test set, against a cross-validated 0.311, and we attributed that gap to a recall ceiling, temporal distribution shift, and threshold miscalibration. We then tested all three. Under leakage-free protocols threshold transfer costs 0.007 F1, decision quality is flat across chronological quartiles, and the official test queries are not measurably farther from the training manifold than training queries are from each other, in two independent embedding spaces. Decomposing the misses instead splits them exactly evenly between candidates never retrieved and candidates retrieved but ranked below the cut. Measuring the remedies for each half, BM25 length-normalisation tuning, an event-triple view, and full-content dense fusion lift top-200 recall by three to seven points, and citation-graph features add 0.014 F1 over eight seeds once own-citation leakage is removed, while per-query cutoff rules, a zero-shot reranker swap, and a date filter do not help. We also document four evaluation artifacts, each of which reversed a result once the protocol was corrected.