uva-irlab-conv at SemEval-2026 Task 8: Multi-Turn RAG with Learned Sparse Retrieval and Listwise Reranking

This report describes our participation in SemEval-2026 Task 8 on multi-turn retrieval and question answering. The task evaluates conversational systems across four domains (finance, cloud documentation, government, Wikipedia), and includes unanswerable queries where the available collection does not contain sufficient evidence to produce a complete response. We propose a multi-turn retrieval-augmented generation pipeline that combines learned sparse retrieval with LLM-based reranking and generation. Using sparse retrieval as the primary retrieval method, we leverage its strong generalization across domains. In addition, we make use of the long-context capabilities of LLMs for conversational query rewriting, pointwise and listwise reranking, and generating the final response, each conditioned on the full conversational history. This multi-step design enables effective integration of conversational context throughout retrieval and generation, improving robustness across domains.

Improving the Efficiency and Effectiveness of LLM Knowledge Distillation for Conversational Search

Conversational Search (CS) considers retrieval of relevant documents based on conversational context. Large Language Models (LLMs) have significantly enhanced CS by enabling effective query rewriting. However, employing LLMs during inference poses efficiency challenges. A method to balance effectiveness and efficiency is the use of knowledge distillation from LLM-based query rewriting. Recent work applies the Kullback-Leibler Divergence (KLD) for distillation, relaxing the alignment with the teacher signal compared to previous methods. Despite these gains, several aspects of KLD-based distillation for conversational search remain understudied, and we investigate them in this work. Prior work in related fields suggests that adding a contrastive loss to the KLD objective can improve performance; we confirm this and observe significant gains in precision-oriented ranking metrics. We also find that contrastive sampling strategies for the KLD loss have a non-trivial impact and must be chosen carefully. Although theory suggests that more samples improve the KLD estimate, experiments show diminishing returns on the number of used samples. Finally, we address the phenomenon of decreased sparsity in longer conversations, which limits computational efficiency across sparse retrieval methods. We find that the representations from the model distilled with the KLD loss can be strongly regularized with a regularization loss, substantially improving sparsity and inference efficiency without significantly harming retrieval effectiveness. We achieve a $2\times$ decrease in FLOPS on TopiOCQA with negligible loss in effectiveness, corresponding to a $\leq 2%$ drop in Recall@100. Our results provide insights into distillation objectives for learned sparse conversational retrievers and offer practical guidelines for improving effectiveness and efficiency in first-stage retrieval.

Lost in Decoding? Reproducing and Stress-Testing the Look-Ahead Prior in Generative Retrieval

Generative retrieval (GR) ranks documents by autoregressively generating document identifiers. Because many GR methods rely on trie-constrained beam search, they are vulnerable to early pruning of relevant prefixes under finite-beam decoding. Planning Ahead in Generative Retrieval (PAG) mitigates this failure mode by using simultaneous decoding to compute a document-level look-ahead prior that guides subsequent sequential decoding. We reproduce PAG at inference time and stress-test its decoding behavior. Using the authors' released checkpoint and identifier/trie artifacts under the reported decoding setup, we reproduce the main effectiveness results on MS MARCO Dev and TREC-DL 2019/2020, and corroborate the reported beam-size-latency trade-off in our hardware setting. Beyond reproduction, we introduce plan drift diagnostics that quantify how intent-preserving query variations alter the planner's top-n candidate set and highest-weight planner tokens, and how these changes affect guided decoding. We find that PAG's planning signal is brittle under lexical surface-form variation: intent-preserving typos can trigger plan collapse, where the planned candidate pool shifts enough that the look-ahead bonus provides little useful guidance, effectively reverting decoding toward weaker unguided search. We further evaluate fixed-index cross-lingual robustness using non-English mMARCO queries against an English index, and assess query-side mitigation strategies that require no re-indexing; query translation provides the strongest recovery in our setting. Overall, our results confirm PAG's reported effectiveness and the benefit of planning-guided decoding under the released inference setup, while showing that these gains depend on the stability of the planning signal under realistic query variation and query-document mismatch.

On the Challenges and Opportunities of Learned Sparse Retrieval for Code

Retrieval over large codebases is a key component of modern LLM-based software engineering systems. Existing approaches predominantly rely on dense embedding models, while learned sparse retrieval (LSR) remains largely unexplored for code. However, applying sparse retrieval to code is challenging due to subword fragmentation, semantic gaps between natural-language queries and code, diversity of programming languages and sub-tasks, and the length of code documents, which can harm sparsity and latency. We introduce SPLADE-Code, the first large-scale family of learned sparse retrieval models specialized for code retrieval (600M-8B parameters). Despite a lightweight one-stage training pipeline, SPLADE-Code achieves state-of-the-art performance among retrievers under 1B parameters (75.4 on MTEB Code) and competitive results at larger scales (79.0 with 8B). We show that learned expansion tokens are critical to bridge lexical and semantic matching, and provide a latency analysis showing that LSR enables sub-millisecond retrieval on a 1M-passage collection with little effectiveness loss.

Unsupervised Corpus Poisoning Attacks in Continuous Space for Dense Retrieval

This paper concerns corpus poisoning attacks in dense information retrieval, where an adversary attempts to compromise the ranking performance of a search algorithm by injecting a small number of maliciously generated documents into the corpus. Our work addresses two limitations in the current literature. First, attacks that perform adversarial gradient-based word substitution search do so in the discrete lexical space, while retrieval itself happens in the continuous embedding space. We thus propose an optimization method that operates in the embedding space directly. Specifically, we train a perturbation model with the objective of maintaining the geometric distance between the original and adversarial document embeddings, while also maximizing the token-level dissimilarity between the original and adversarial documents. Second, it is common for related work to have a strong assumption that the adversary has prior knowledge about the queries. In this paper, we focus on a more challenging variant of the problem where the adversary assumes no prior knowledge about the query distribution (hence, unsupervised). Our core contribution is an adversarial corpus attack that is fast and effective. We present comprehensive experimental results on both in- and out-of-domain datasets, focusing on two related tasks: a top-1 attack and a corpus poisoning attack. We consider attacks under both a white-box and a black-box setting. Notably, our method can generate successful adversarial examples in under two minutes per target document; four times faster compared to the fastest gradient-based word substitution methods in the literature with the same hardware. Furthermore, our adversarial generation method generates text that is more likely to occur under the distribution of natural text (low perplexity), and is therefore more difficult to detect.

Conversational Gold: Evaluating Personalized Conversational Search System using Gold Nuggets

The rise of personalized conversational search systems has been driven by advancements in Large Language Models (LLMs), enabling these systems to retrieve and generate answers for complex information needs. However, the automatic evaluation of responses generated by Retrieval Augmented Generation (RAG) systems remains an understudied challenge. In this paper, we introduce a new resource for assessing the retrieval effectiveness and relevance of response generated by RAG systems, using a nugget-based evaluation framework. Built upon the foundation of TREC iKAT 2023, our dataset extends to the TREC iKAT 2024 collection, which includes 17 conversations and 20,575 relevance passage assessments, together with 2,279 extracted gold nuggets, and 62 manually written gold answers from NIST assessors. While maintaining the core structure of its predecessor, this new collection enables a deeper exploration of generation tasks in conversational settings. Key improvements in iKAT 2024 include: (1) ``gold nuggets'' -- concise, essential pieces of information extracted from relevant passages of the collection -- which serve as a foundation for automatic response evaluation; (2) manually written answers to provide a gold standard for response evaluation; (3) unanswerable questions to evaluate model hallucination; (4) expanded user personas, providing richer contextual grounding; and (5) a transition from Personal Text Knowledge Base (PTKB) ranking to PTKB classification and selection. Built on this resource, we provide a framework for long-form answer generation evaluation, involving nuggets extraction and nuggets matching, linked to retrieval. This establishes a solid resource for advancing research in personalized conversational search and long-form answer generation. Our resources are publicly available at https://github.com/irlabamsterdam/CONE-RAG.

Reproducing NevIR: Negation in Neural Information Retrieval

Negation is a fundamental aspect of human communication, yet it remains a challenge for Language Models (LMs) in Information Retrieval (IR). Despite the heavy reliance of modern neural IR systems on LMs, little attention has been given to their handling of negation. In this study, we reproduce and extend the findings of NevIR, a benchmark study that revealed most IR models perform at or below the level of random ranking when dealing with negation. We replicate NevIR's original experiments and evaluate newly developed state-of-the-art IR models. Our findings show that a recently emerging category - listwise Large Language Model (LLM) rerankers - outperforms other models but still underperforms human performance. Additionally, we leverage ExcluIR, a benchmark dataset designed for exclusionary queries with extensive negation, to assess the generalizability of negation understanding. Our findings suggest that fine-tuning on one dataset does not reliably improve performance on the other, indicating notable differences in their data distributions. Furthermore, we observe that only cross-encoders and listwise LLM rerankers achieve reasonable performance across both negation tasks.

IRLab@iKAT24: Learned Sparse Retrieval with Multi-aspect LLM Query Generation for Conversational Search

The Interactive Knowledge Assistant Track (iKAT) 2024 focuses on advancing conversational assistants, able to adapt their interaction and responses from personalized user knowledge. The track incorporates a Personal Textual Knowledge Base (PTKB) alongside Conversational AI tasks, such as passage ranking and response generation. Query Rewrite being an effective approach for resolving conversational context, we explore Large Language Models (LLMs), as query rewriters. Specifically, our submitted runs explore multi-aspect query generation using the MQ4CS framework, which we further enhance with Learned Sparse Retrieval via the SPLADE architecture, coupled with robust cross-encoder models. We also propose an alternative to the previous interleaving strategy, aggregating multiple aspects during the reranking phase. Our findings indicate that multi-aspect query generation is effective in enhancing performance when integrated with advanced retrieval and reranking models. Our results also lead the way for better personalization in Conversational Search, relying on LLMs to integrate personalization within query rewrite, and outperforming human rewrite performance.

DiSCo Meets LLMs: A Unified Approach for Sparse Retrieval and Contextual Distillation in Conversational Search

Conversational Search (CS) is the task of retrieving relevant documents from a corpus within a conversational context, combining retrieval with conversational context modeling. With the explosion of Large Language Models (LLMs), the CS field has seen major improvements with LLMs rewriting user queries, accounting for conversational context. However, engaging LLMs at inference time harms efficiency. Current methods address this by distilling embeddings from human-rewritten queries to learn the context modeling task. Yet, these approaches predominantly focus on context modeling, and only treat the contrastive component of the retrieval task within a distillation-independent loss term. To address these limitations, we propose a new distillation method, as a relaxation of the previous objective, unifying retrieval and context modeling. We relax the existing training objectives by distilling similarity scores between conversations and documents, rather than relying solely on representation learning. Our proposed distillation objective allows for more freedom in the representation space and leverages the contrastive nature of document relevance. Through experiments on Learned Sparse Retrieval (LSR) across 5 CS datasets, our approach demonstrates substantial improvements in both in-domain and out-of-domain retrieval performance, outperforming state-of-the-art with gains of up to 6 points in recall for out-of-domain datasets. Additionally, through the relaxation of the objective, we propose a multi-teacher distillation, using multiple LLMs as teachers, yielding additional gains, and outperforming the teachers themselves in in-domain experiments. Finally, analysis of the sparsity of the models reveals that our distillation allows for better control over the sparsity of the trained models.

Benchmarking Middle-Trained Language Models for Neural Search

Middle training methods aim to bridge the gap between the Masked Language Model (MLM) pre-training and the final finetuning for retrieval. Recent models such as CoCondenser, RetroMAE, and LexMAE argue that the MLM task is not sufficient enough to pre-train a transformer network for retrieval and hence propose various tasks to do so. Intrigued by those novel methods, we noticed that all these models used different finetuning protocols, making it hard to assess the benefits of middle training. We propose in this paper a benchmark of CoCondenser, RetroMAE, and LexMAE, under the same finetuning conditions. We compare both dense and sparse approaches under various finetuning protocols and middle training on different collections (MS MARCO, Wikipedia or Tripclick). We use additional middle training baselines, such as a standard MLM finetuning on the retrieval collection, optionally augmented by a CLS predicting the passage term frequency. For the sparse approach, our study reveals that there is almost no statistical difference between those methods: the more effective the finetuning procedure is, the less difference there is between those models. For the dense approach, RetroMAE using MS MARCO as middle-training collection shows excellent results in almost all the settings. Finally, we show that middle training on the retrieval collection, thus adapting the language model to it, is a critical factor. Overall, a better experimental setup should be adopted to evaluate middle training methods. Code available at https://github.com/naver/splade/tree/benchmarch-SIGIR23