| Challenge: | Retrieval systems rely on surface-level cues such as keyword overlap and semantic similarity to evaluate retrieval beyond these shallow signals. |
| Approach: | They propose a benchmark that shifts the reasoning challenge to query-side processing techniques that can help resolve complexity. |
| Outcome: | The proposed benchmarks show that document-side reasoning remains a challenge. |
Similar Papers
UniRAG: A Unified RAG Framework for Knowledge-Intensive Queries with Decomposition, Break-Down Reasoning, and Iterative Rewriting (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing retrieval-augmented generation approaches struggle with query complexity, propagated reasoning errors, or rely on incomplete or noisy retrieval. |
| Approach: | a unified retrieval-augmented generation framework is developed to address query complexity . the framework decomposes queries into semantically coherent sub-queries . it explicitly verifies retrieved sub-facts and adaptively refines queries based on identified knowledge gaps. |
| Outcome: | a new framework improves answer completeness and reliability by decomposing queries into coherent sub-queries . the framework explicitly verifies retrieved sub-facts and adaptively refines queries based on identified knowledge gaps. |
Benchmarking Query-Conditioned Natural Language Inference (2025.findings-acl)
Copied to clipboard
| Challenge: | Query-conditioned natural language inference (QC-NLI) is a new approach to detect inconsistencies in large language models. |
| Approach: | They propose a task of Query-Conditioned Natural Language Inference to determine the semantic relationship between two documents conditioned on a query. |
| Outcome: | The proposed task is based on a query-conditioned natural language inference (QC-NLI) it is used to determine the relationship between the premise and hypothesis given a given query. |
Logic Haystacks: Probing LLMs’ Long-Context Logical Reasoning (Without Easily Identifiable Unrelated Padding) (2026.eacl-short)
Copied to clipboard
| Challenge: | Recent large language models claim long context windows, but evaluations often involve simple retrieval tasks or synthetic tasks padded with irrelevant text. |
| Approach: | They use grammars to generate simplified English with logical representations to create long input text while controlling its semantics. |
| Outcome: | The proposed model performs better with realistic distractors than with standard models. |
SCROLLS: Standardized CompaRison Over Long Language Sequences (2022.emnlp-main)
Copied to clipboard
Uri Shaham, Elad Segal, Maor Ivgi, Avia Efrat, Ori Yoran, Adi Haviv, Ankit Gupta, Wenhan Xiong, Mor Geva, Jonathan Berant, Omer Levy
| Challenge: | Standard NLP benchmarks focus on short texts, but long texts are produced in the context of longer discourses. |
| Approach: | They propose a new benchmark that places models in context of long texts that require reasoning over long texts. |
| Outcome: | The proposed task sets are based on a set of long-text datasets and host a live leaderboard to facilitate research on model architecture and pretraining methods. |
Assessing “Implicit” Retrieval Robustness of Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Retrieval-augmented generation (RAG) is a framework to enhance large language models with external knowledge, but its effectiveness is constrained by the retrieval robustness of the model. |
| Approach: | They propose to use gold and distracting context to fine-tune models to handle relevant or irrelevant retrieved context in an end-to-end manner. |
| Outcome: | The proposed model performs better when gold and distracting context are used, while still extracting correct answers when retrieval is accurate. |
One Thousand and One Pairs: A “novel” challenge for long-context language models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing long-context evaluation methods measure surface-level retrieval capabilities, but do not assess performance on the more challenging task of synthesizing distant and underlying information. |
| Approach: | They propose a dataset of 1,001 minimally different pairs of true and false claims about 67 recently-published English fictional books. |
| Outcome: | The proposed model performs better on pairs that require only sentence-level retrieval vs. global reasoning . the proposed model also performs worse on speculative fiction books with extensive world-building . |
Can LLMs reason over extended multilingual contexts? Towards long-context evaluation beyond retrieval over haystacks (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing multilingual long-context benchmarks are myopic and inherently limited, as successful recall alone does not indicate a model’s capacity to reason over extended contexts. |
| Approach: | They propose a new synthetic benchmark for multilingual long-context reasoning that includes bAbI-style tasks that test multi-hop inference, aggregation, and epistemic reasoning. |
| Outcome: | The proposed benchmarks are based on a multilingual long-context model and span seven languages. |
Benchmarking Deflection and Hallucination in Large Vision-Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks overlook conflicts between visual and textual evidence and the importance of generating deflections when incomplete knowledge is retrieved. |
| Approach: | They propose a dynamic curation pipeline that preserves benchmark difficulty over time . they propose 'vlm-DeflectionBench' benchmark to probe model behaviour under conflicting evidence . |
| Outcome: | The proposed benchmarks overlook conflicts between visual and textual evidence and are prone to obsolescence . the proposed benchmark is based on 2,775 samples spanning diverse retrieval settings . |
WikiHowQA: A Comprehensive Benchmark for Multi-Document Non-Factoid Question Answering (2023.acl-long)
Copied to clipboard
| Challenge: | Answering non-factoid questions (NFQs) is a challenging task, requiring passage-level answers that are difficult to construct and evaluate. |
| Approach: | They propose a multi-document NFQA benchmark built on WikiHow, a website dedicated to answering “how-to” questions. |
| Outcome: | The proposed framework includes 11,746 human-written answers along with 74,527 supporting documents. |
Collapse of Dense Retrievers: Short, Early, and Literal Biases Outranking Factual Evidence (2025.acl-long)
Copied to clipboard
| Challenge: | Notably, when multiple biases combine, models exhibit catastrophic performance degradation, selecting the answer-containing document in less than 10% of cases over a synthetic biased document without the answer. |
| Approach: | They repurpose a relation extraction dataset to quantify the impact of heuristic biases on retrievers like Dragon+ and Contriever. |
| Outcome: | The proposed models exhibit catastrophic performance degradation when multiple biases combine, selecting the answer-containing document in less than 10% of cases over a synthetic biased document without the answer. |