Jaydeep Sen

Avirup Sil, Jaydeep Sen, Bhavani Iyer et al. · ibm-research

The field of Question Answering (QA) has made remarkable progress in recent years, thanks to the advent of large pre-trained language models, newer realistic benchmark datasets with leaderboards, and novel algorithms for key components such as retrievers and readers. In this paper, we introduce PRIMEQA: a one-stop and open-source QA repository with an aim to democratize QA re-search and facilitate easy replication of state-of-the-art (SOTA) QA methods. PRIMEQA supports core QA functionalities like retrieval and reading comprehension as well as auxiliary capabilities such as question generation.It has been designed as an end-to-end toolkit for various use cases: building front-end applications, replicating SOTA methods on pub-lic benchmarks, and expanding pre-existing methods. PRIMEQA is available at : https://github.com/primeqa.

Parul Awasthy, Aashka Trivedi, Yushu Yang et al.

We introduce the multilingual Granite Embedding R2 models, a family of encoder-based embedding models for enterprise-scale dense retrieval across 200+ languages. Extending our English-focused R2 release, these models add enhanced support for 52 languages and programming code, a 32,768-token context window (a 64x expansion over R1), and state-of-the-art overall performance across multilingual and cross-lingual text search, code retrieval, long-document search, and reasoning retrieval datasets. The release consists of two bi-encoder models based on the ModernBERT architecture with an expanded multilingual vocabulary: a 311M-parameter full-size, and a 97M-parameter compact model built via model pruning and vocabulary selection that achieves the highest retrieval score of any open multilingual embedding model under 100M parameters. The full-size also supports Matryoshka Representation Learning for flexible embedding dimensionality. Both models are trained on enterprise-appropriate data with governance oversight, and released under the Apache 2.0 license at https://huggingface.co/collections/ibm-granite, designed to support responsible use and enable unrestricted research and enterprise adoption.

Abhishek Kumar Singh, Vishwajeet kumar, Rudra Murthy et al.

Large Language Models (LLMs) perform well on unseen tasks in English, but their abilities in non English languages are less explored due to limited benchmarks and training data. To bridge this gap, we introduce the Indic QA Benchmark, a large dataset for context grounded question answering in 11 major Indian languages, covering both extractive and abstractive tasks. Evaluations of multilingual LLMs, including instruction finetuned versions, revealed weak performance in low resource languages due to a strong English language bias in their training data. We also investigated the Translate Test paradigm,where inputs are translated to English for processing and the results are translated back into the source language for output. This approach outperformed multilingual LLMs, particularly in low resource settings. By releasing Indic QA, we aim to promote further research into LLMs question answering capabilities in low resource languages. This benchmark offers a critical resource to address existing limitations and foster multilingual understanding.

Parul Awasthy, Aashka Trivedi, Yulong Li et al. · ibm-research

We introduce the Granite Embedding models, a family of encoder-based embedding models designed for retrieval tasks, spanning dense-retrieval and sparse retrieval architectures, with both English and Multilingual capabilities. This report provides the technical details of training these highly effective 12 layer embedding models, along with their efficient 6 layer distilled counterparts. Extensive evaluations show that the models, developed with techniques like retrieval oriented pretraining, contrastive finetuning, knowledge distillation, and model merging significantly outperform publicly available models of similar sizes on both internal IBM retrieval and search tasks, and have equivalent performance on widely used information retrieval benchmarks, while being trained on high-quality data suitable for enterprise use. We publicly release all our Granite Embedding models under the Apache 2.0 license, allowing both research and commercial use at https://huggingface.co/collections/ibm-granite.

Meet Doshi, Vishwajeet Kumar, Rudra Murthy et al.

Learned Sparse Retrievers (LSR) have evolved into an effective retrieval strategy that can bridge the gap between traditional keyword-based sparse retrievers and embedding-based dense retrievers. At its core, learned sparse retrievers try to learn the most important semantic keyword expansions from a query and/or document which can facilitate better retrieval with overlapping keyword expansions. LSR like SPLADE has typically been using encoder only models with MLM (masked language modeling) style objective in conjunction with known ways of retrieval performance improvement such as hard negative mining, distillation, etc. In this work, we propose to use decoder-only model for learning semantic keyword expansion. We posit, decoder only models that have seen much higher magnitudes of data are better equipped to learn keyword expansions needed for improved retrieval. We use Mistral as the backbone to develop our Learned Sparse Retriever similar to SPLADE and train it on a subset of sentence-transformer data which is often used for training text embedding models. Our experiments support the hypothesis that a sparse retrieval model based on decoder only large language model (LLM) surpasses the performance of existing LSR systems, including SPLADE and all its variants. The LLM based model (Echo-Mistral-SPLADE) now stands as a state-of-the-art learned sparse retrieval model on the BEIR text retrieval benchmark.

Parul Awasthy, Aashka Trivedi, Yulong Li et al. · ibm-research

We introduce the Granite Embedding R2 models, a comprehensive family of high-performance English encoder-based embedding models engineered for enterprise-scale dense retrieval applications. Building upon our first-generation release, these models deliver substantial improvements, including 16x expanded context length (8,192 tokens), state-of-the-art performance across diverse retrieval domains - text, code, long-document search, multi-turn conversational, and tabular data - and measurable speed advantages of 19-44\% over leading competitors while maintaining superior accuracy. Our release encompasses both bi-encoder and cross-encoder architectures, featuring a highly effective 22-layer retriever model and its efficient 12-layer counterpart, alongside a high-quality reranker model, all trained exclusively on enterprise-appropriate data with comprehensive governance oversight. The models demonstrate exceptional versatility across standard benchmarks, IBM-developed evaluation suites, and real-world enterprise use cases, establishing new performance standards for open-source embedding models. In an era where retrieval speed and accuracy are paramount for competitive advantage, the Granite R2 models deliver a compelling combination of cutting-edge performance, enterprise-ready licensing, and transparent data provenance that organizations require for mission-critical deployments. All models are publicly available under the Apache 2.0 license at https://huggingface.co/collections/ibm-granite, enabling unrestricted research and commercial use.

Arkadeep Acharya, Rudra Murthy, Vishwajeet Kumar et al.

Given the large number of Hindi speakers worldwide, there is a pressing need for robust and efficient information retrieval systems for Hindi. Despite ongoing research, comprehensive benchmarks for evaluating retrieval models in Hindi are lacking. To address this gap, we introduce the Hindi-BEIR benchmark, comprising 15 datasets across seven distinct tasks. We evaluate state-of-the-art multilingual retrieval models on the Hindi-BEIR benchmark, identifying task and domain-specific challenges that impact Hindi retrieval performance. Building on the insights from these results, we introduce NLLB-E5, a multilingual retrieval model that leverages a zero-shot approach to support Hindi without the need for Hindi training data. We believe our contributions, which include the release of the Hindi-BEIR benchmark and the NLLB-E5 model, will prove to be a valuable resource for researchers and promote advancements in multilingual retrieval models.

Arkadeep Acharya, Rudra Murthy, Vishwajeet Kumar et al.

Given the large number of Hindi speakers worldwide, there is a pressing need for robust and efficient information retrieval systems for Hindi. Despite ongoing research, there is a lack of comprehensive benchmark for evaluating retrieval models in Hindi. To address this gap, we introduce the Hindi version of the BEIR benchmark, which includes a subset of English BEIR datasets translated to Hindi, existing Hindi retrieval datasets, and synthetically created datasets for retrieval. The benchmark is comprised of $15$ datasets spanning across $8$ distinct tasks. We evaluate state-of-the-art multilingual retrieval models on this benchmark to identify task and domain-specific challenges and their impact on retrieval performance. By releasing this benchmark and a set of relevant baselines, we enable researchers to understand the limitations and capabilities of current Hindi retrieval models, promoting advancements in this critical area. The datasets from Hindi-BEIR are publicly available.

Meet Doshi, Aashka Trivedi, Vishwajeet Kumar et al.

Representation learning is central to many downstream tasks such as search, clustering, classification, and reranking. State-of-the-art sequence encoders typically collapse a variable-length token sequence to a single vector using a pooling operator, most commonly a special [CLS] token or mean pooling over token embeddings. In this paper, we identify systematic weaknesses of these pooling strategies: [CLS] tends to concentrate information toward the initial positions of the sequence and can under-represent distributed evidence, while mean pooling can dilute salient local signals, sometimes leading to worse short-context performance. To address these issues, we introduce Landmark (LMK) pooling, which partitions a sequence into chunks, inserts landmark tokens between chunks, and forms the final representation by mean-pooling the landmark token embeddings. This simple mechanism improves long-context extrapolation without sacrificing local salient features, at the cost of introducing a small number of special tokens. We empirically demonstrate that LMK pooling matches existing methods on short-context retrieval tasks and yields substantial improvements on long-context tasks, making it a practical and scalable alternative to existing pooling methods.

Meet Doshi, Vishwajeet Kumar, Yulong Li et al.

General-purpose open-domain dense retrieval systems are usually trained with a large, eclectic mix of corpora and search tasks. How should these diverse corpora and tasks be sampled for training? Conventional approaches sample them uniformly, proportional to their instance population sizes, or depend on human-level expert supervision. It is well known that the training data sampling strategy can greatly impact model performance. However, how to find the optimal strategy has not been adequately studied in the context of embedding models. We propose Inf-DDS, a novel reinforcement learning driven sampling framework that adaptively reweighs training datasets guided by influence-based reward signals and is much more lightweight with respect to GPU consumption. Our technique iteratively refines the sampling policy, prioritizing datasets that maximize model performance on a target development set. We evaluate the efficacy of our sampling strategy on a wide range of text retrieval tasks, demonstrating strong improvements in retrieval performance and better adaptation compared to existing gradient-based sampling methods, while also being 1.5x to 4x cheaper in GPU compute. Our sampling strategy achieves a 5.03 absolute NDCG@10 improvement while training a multilingual bge-m3 model and an absolute NDCG@10 improvement of 0.94 while training all-MiniLM-L6-v2, even when starting from expert-assigned weights on a large pool of training datasets.

Sshubam Verma, Mohammed Safi Ur Rahman Khan, Vishwajeet Kumar et al.

Evaluating Large Language Models (LLMs) in low-resource and linguistically diverse languages remains a significant challenge in NLP, particularly for languages using non-Latin scripts like those spoken in India. Existing benchmarks predominantly focus on English, leaving substantial gaps in assessing LLM capabilities in these languages. We introduce MILU, a Multi task Indic Language Understanding Benchmark, a comprehensive evaluation benchmark designed to address this gap. MILU spans 8 domains and 41 subjects across 11 Indic languages, reflecting both general and culturally specific knowledge. With an India-centric design, incorporates material from regional and state-level examinations, covering topics such as local history, arts, festivals, and laws, alongside standard subjects like science and mathematics. We evaluate over 42 LLMs, and find that current LLMs struggle with MILU, with GPT-4o achieving the highest average accuracy at 74 percent. Open multilingual models outperform language-specific fine-tuned models, which perform only slightly better than random baselines. Models also perform better in high resource languages as compared to low resource ones. Domain-wise analysis indicates that models perform poorly in culturally relevant areas like Arts and Humanities, Law and Governance compared to general fields like STEM. To the best of our knowledge, MILU is the first of its kind benchmark focused on Indic languages, serving as a crucial step towards comprehensive cultural evaluation. All code, benchmarks, and artifacts are publicly available to foster open research.

Vishwajeet Kumar, Yash Gupta, Saneem Chemmengath et al.

Question answering (QA) over tables and linked text, also called TextTableQA, has witnessed significant research in recent years, as tables are often found embedded in documents along with related text. HybridQA and OTT-QA are the two best-known TextTableQA datasets, with questions that are best answered by combining information from both table cells and linked text passages. A common challenge in both datasets, and TextTableQA in general, is that the training instances include just the question and answer, where the gold answer may match not only multiple table cells across table rows but also multiple text spans within the scope of a table row and its associated text. This leads to a noisy multi instance training regime. We present MITQA, a transformer-based TextTableQA system that is explicitly designed to cope with distant supervision along both these axes, through a multi-instance loss objective, together with careful curriculum design. Our experiments show that the proposed multi-instance distant supervision approach helps MITQA get state-of-the-art results beating the existing baselines for both HybridQA and OTT-QA, putting MITQA at the top of HybridQA leaderboard with best EM and F1 scores on a held out test set.

Saneem Ahmed Chemmengath, Vishwajeet Kumar, Samarth Bharadwaj et al.

Weakly-supervised table question-answering(TableQA) models have achieved state-of-art performance by using pre-trained BERT transformer to jointly encoding a question and a table to produce structured query for the question. However, in practical settings TableQA systems are deployed over table corpora having topic and word distributions quite distinct from BERT's pretraining corpus. In this work we simulate the practical topic shift scenario by designing novel challenge benchmarks WikiSQL-TS and WikiTQ-TS, consisting of train-dev-test splits in five distinct topic groups, based on the popular WikiSQL and WikiTableQuestions datasets. We empirically show that, despite pre-training on large open-domain text, performance of models degrades significantly when they are evaluated on unseen topics. In response, we propose T3QA (Topic Transferable Table Question Answering) a pragmatic adaptation framework for TableQA comprising of: (1) topic-specific vocabulary injection into BERT, (2) a novel text-to-text transformer generator (such as T5, GPT2) based natural language question generation pipeline focused on generating topic specific training data, and (3) a logical form reranker. We show that T3QA provides a reasonably good baseline for our topic shift benchmarks. We believe our topic split benchmarks will lead to robust TableQA solutions that are better suited for practical deployment.

Yannis Katsis, Saneem Chemmengath, Vishwajeet Kumar et al.

Recent advances in transformers have enabled Table Question Answering (Table QA) systems to achieve high accuracy and SOTA results on open domain datasets like WikiTableQuestions and WikiSQL. Such transformers are frequently pre-trained on open-domain content such as Wikipedia, where they effectively encode questions and corresponding tables from Wikipedia as seen in Table QA dataset. However, web tables in Wikipedia are notably flat in their layout, with the first row as the sole column header. The layout lends to a relational view of tables where each row is a tuple. Whereas, tables in domain-specific business or scientific documents often have a much more complex layout, including hierarchical row and column headers, in addition to having specialized vocabulary terms from that domain. To address this problem, we introduce the domain-specific Table QA dataset AIT-QA (Airline Industry Table QA). The dataset consists of 515 questions authored by human annotators on 116 tables extracted from public U.S. SEC filings (publicly available at: https://www.sec.gov/edgar.shtml) of major airline companies for the fiscal years 2017-2019. We also provide annotations pertaining to the nature of questions, marking those that require hierarchical headers, domain-specific terminology, and paraphrased forms. Our zero-shot baseline evaluation of three transformer-based SOTA Table QA methods - TaPAS (end-to-end), TaBERT (semantic parsing-based), and RCI (row-column encoding-based) - clearly exposes the limitation of these methods in this practical setting, with the best accuracy at just 51.8\% (RCI). We also present pragmatic table preprocessing steps used to pivot and project these complex tables into a layout suitable for the SOTA Table QA models.