arXiv subjects

Dipankar Srirag

Publications and source records attributed to Dipankar Srirag.

14 recordsLinked to original sources

arXiv2026

LLMs Anchor on Chief Complaint and Fail to Integrate Evidence in Sequential Clinical Triage

Triage in the emergency department (ED) is a sequential decision process that unfolds turn by turn. Existing evaluations of large language models (LLMs) for triage use completed retrospective records and report performance close to that of physicians. We implement a methodology for evaluating LLMs on sequential triage, the task of predicting a triage acuity label from a growing prefix of a nurse-patient conversation. We evaluate six LLMs at five sequential checkpoints on two corpora: 425 LLM-generated (SIMULATED) and 50 physician-authored (CLINICIAN) conversations, both labelled under the Emergency Severity Index (ESI). Every model, measured by quadratic weighted kappa (QWK), degrades from moderate-to-substantial agreement on completed records to fair-to-moderate agreement at every sequential checkpoint. Controlled perturbations show that the label at every checkpoint is anchored on the chief complaint exchanges, and prompting interventions fail to lift this plateau. Models extract clinically relevant content from later turns, yet the surprisal of the true label rises across the checkpoints. So the model fails to integrate the evidence. Three expert clinicians on the same conversations reach a QWK of 0.887-0.929, while the best model reaches 0.295. Predictions concentrate at ESI-2 and ESI-3, and models agree with each other more than with the ground truth, so ensembling worsens the failure. Deploying LLMs for ED triage based on offline benchmarks alone misses this sequential failure.

cs.CL↗

arXiv2026

DiaLLM: An Investigation into the Robustness-Generation Gap in English Dialect Adaptation

Large language models increasingly understand dialectal English, yet still produce only standard, US-leaning English, leaving dialectal generation, the harder half of the problem, largely unaddressed. We introduce DiaLLM, which continually pretrains three open-weight language model families on the International Corpus of English and applies implicit and explicit post-training paradigms, each combined with three model alignment strategies, giving the first controlled comparison of these components across Australian, Indian, and Northern British English. Our results reveal a robustness-generation gap: benchmarks are shaped by continual pretraining and SFT, while alignment visibly reshapes generation in ways benchmarks do not capture. Explicit variety-targeted adaptation produces output reliably recognised as dialectal and judged more dialectal than broad alignment, yet where human judgement was directly assessed, the method that most aggressively optimises the dialectal reward is not the one judged most dialectal. Independent linguistic analysis corroborates this reward-quality gap, most clearly on two of the three families. No single alignment method dominates, and closing the gap will require richer reward designs and continued investment in dialectal resources. We release all code, checkpoints, and preference datasets.

cs.CL↗

arXiv2026

Last Translation Benchmark

For scientific progress, we need benchmarks that test the limits of state-of-the-art models, and evaluation methods that inform us about failure cases. As models get stronger, standard benchmarks for machine translation are approaching saturation. Further, automatic translation metrics are unreliable, vulnerable to reward-hacking, and provide unactionable assessments. Even gold human evaluation is not problem-free, because it often lacks reproducibility, objectivity, and scalability. Overall, this prevents us from tracking objective progress in the field and identifying pathways for improvement. We introduce the Last Translation Benchmark, a collection of human-authored and peer-reviewed examples (texts, images, audio, videos) that break leading machine translation models. We also present a new evaluation approach: each example comes with handcrafted verification rules describing concrete failure cases on that example, therefore allowing reliable and actionable future evaluation. The Last Translation Benchmark is a live dataset that accepts ongoing contributions. The latest version is LTBv1, containing accepted contributions prior to September 1st 2026, with future releases planned as new data is continuously collected.

cs.CL↗

arXiv2026

From Triage to Discharge: A Survey of NLP Tasks, Methods, and Open Challenges in the Emergency Department

Emergency departments (EDs) operate under time pressure, generating multimodal data such as clinical conversations, triage notes, and discharge documents. Recent advances in natural language processing (NLP), particularly pretrained transformers and large language models, have created new opportunities to support language and time-intensive stages of emergency care. Yet existing surveys map clinical NLP across the broader hospital workflow or focus on specific tasks. This survey analyses 46 papers spanning the three phases of ED: triage, diagnosis, and disposition, covering tasks such as triage classification, clinical summarisation, automatic diagnosis, report generation, and discharge documentation. We examine modelling paradigms, evaluation practices, and emerging benchmarks and shared tasks. Across tasks, we identify common trends, including a shift from task-specific neural architectures to pretrained language models, growing interest in interactive clinical systems, and increasing attention to clinically grounded evaluation. Finally, we detail open challenges such as limited generalisability, noisy clinical inputs, and workflow constraints that inform future ED-NLP research.

cs.CL↗

arXiv2026

TriageSim: A Conversational Emergency Triage Simulation Framework from Structured Electronic Health Records

Research in emergency triage is restricted to structured electronic health records (EHR) due to regulatory constraints on nurse-patient interactions. We introduce TriageSim, a simulation framework for generating persona-conditioned triage conversations from structured records. TriageSim enables multi-turn nurse-patient interactions with explicit control over disfluency and decision behaviour, producing a corpus of ~800 synthetic transcripts and corresponding audio. We use a combination of automated analysis for linguistic, behavioural and acoustic fidelity alongside manual evaluation for medical fidelity using a random subset of 50 conversations. The utility of the generated corpus is examined via conversational triage classification. We observe modest agreement for acuity levels across three modalities: generated synthetic text, ASR transcripts, and direct audio inputs. We provide the code for TriageSim at https://github.com/dipankarsrirag/triage-sim.git.

cs.CL↗

arXiv2026

Retell, Reward, Repeat: Reinforcement Learning for Narrative Theory-Informed Story Retelling

Counterfactual story retelling exposes LLM shortcomings in constrained narrative solution spaces where they can no longer rely on recalling memorised training data. Ground-truth-based post-training, such as SFT, fails to teach LLMs how to generate logical and rational narrative events. In this paper, we introduce Retell, Reward, Repeat (RRR), an RL-based pipeline synthesising Structuralist Narratology with scalar narrativity to teach storytelling structure. We extend the TimeTravel dataset with human-annotated stages of narrative equilibrium to evaluate reward models. By using d-RLAIF, RRR derives training signals from the narrativity of textual features without the need for reference outputs. Evaluations demonstrate that RRR-trained LLMs outperform few-shot and SFT baselines in logic, rationality, and completeness, with output quality additionally validated by blind human preference. Relying on a small, query-only dataset, RRR provides a linguistically grounded, cost-effective post-training mechanism for storytelling--a domain currently lacking effective post-training methods. RRR highlights the continued relevance of integrating established linguistic theories into contemporary NLP.

cs.CL↗

arXiv2026

PAREDA: A Multi-Accent Speech Dataset of Natural Language Processing Research Discussions

While modern Automatic Speech Recognition (ASR) systems achieve high accuracy on benchmark corpora, their performance often degrades when there is real-world variability. This work focuses on variability arising due to accented, spontaneous, and domain-specific speech. In particular, we introduce PAper REading DAtaset (PAREDA), a first-of-its-kind multi-accent speech dataset consisting of discussions on academic Natural Language Processing (NLP) papers between speakers with Australian, Indian-English, and Chinese English accents. Each session elicits a spontaneous monologue (a summary of a paper's abstract) and a non-monologue (a question-and-answer session between participants), resulting in a corpus rich with technical jargon and conversational phenomena. We evaluate the performance of SOTA ASR models on PAREDA, analysing the impact of accent mixing and increased speech rate. Our results show that, in the zero-shot setting, models perform worse, confirming the dataset's challenging nature. However, fine-tuning on PAREDA significantly reduces the Word Error Rate (WER), demonstrating that our dataset captures linguistic characteristics often missing from existing corpora. PAREDA serves as a valuable new resource for building and evaluating more robust and inclusive ASR systems for specialised, real-world applications.

cs.CL↗

arXiv2026

Far Out: Evaluating Language Models on Slang in Australian and Indian English

Language models exhibit systematic performance gaps when processing text in non-standard language varieties, yet their ability to comprehend variety-specific slang remains underexplored for several languages. We present a comprehensive evaluation of slang awareness in Indian English (en-IN) and Australian English (en-AU) across seven state-of-the-art language models. We construct two complementary datasets: WEB, containing 377 web-sourced usage examples from Urban Dictionary, and GEN, featuring 1,492 synthetically generated usages of these slang terms, across diverse scenarios. We assess language models on three tasks: target word prediction (TWP), guided target word prediction (TWP$^*$) and target word selection (TWS). Our results reveal four key findings: (1) Higher average model performance TWS versus TWP and TWP$^*$, with average accuracy score increasing from 0.03 to 0.49 respectively (2) Stronger average model performance on WEB versus GEN datasets, with average similarity score increasing by 0.03 and 0.05 across TWP and TWP$^*$ tasks respectively (3) en-IN tasks outperform en-AU when averaged across all models and datasets, with TWS demonstrating the largest disparity, increasing average accuracy from 0.44 to 0.54. These findings underscore fundamental asymmetries between generative and discriminative competencies for variety-specific language, particularly in the context of slang expressions despite being in a technologically rich language such as English.

cs.CL↗

arXiv2025

A Taxonomy-Driven Case Study of Australian Web Resources Against Technology-Facilitated Abuse

Technology-Facilitated Abuse (TFA) encompasses a broad and rapidly evolving set of behaviours in which digital systems are used to harass, monitor, threaten, or control individuals. Although prior research has documented many forms of TFA, there is no consolidated framework for understanding how abuse types, prevention measures, detection mechanisms, and support pathways relate across the abuse life cycle. This paper contributes a unified, literature-derived taxonomy of TFA grounded in a structured review of peer-reviewed studies, and the first large-scale, taxonomy-aligned audit of institutional web resources in Australia. We crawl 306 government, non-government, and service-provider domains, obtaining 52,605 pages, and classify using zero-shot topic models to map web content onto our taxonomy. An emotion and readability analyses reveal how institutions frame TFA and how accessible their guidance is to the public. Our findings show that institutional websites cover only a narrow subset of harms emphasised in the literature, with approximately 70% of all abuse labelled pages focused on harassment, comments abuse, or sexual abuse, while less than 1% address covert surveillance, economic abuse, or long-term controlling behaviours. Support pathways are similarly limited, with most resources centred on digital information hubs rather than counselling or community-based services. Readability analysis further shows that much of this content is written at late secondary or early tertiary reading levels, which may be inaccessible to a substantial portion of at-risk users. By highlighting strengths and gaps in Australia's support for TFA, our taxonomy and audit method provide a scalable basis for evaluating institutional communication, improving survivor resources, and guiding safer digital ecosystems. The taxonomy serves as a foundation for analyses in national contexts to foster TFA awareness.

cs.CY↗

arXiv2025

Nek Minit: Harnessing Pragmatic Metacognitive Prompting for Explainable Sarcasm Detection of Australian and Indian English

Sarcasm is a challenge to sentiment analysis because of the incongruity between stated and implied sentiment. The challenge is exacerbated when the implication may be relevant to a specific country or geographical region. Pragmatic metacognitive prompting (PMP) is a cognition-inspired technique that has been used for pragmatic reasoning. In this paper, we harness PMP for explainable sarcasm detection for Australian and Indian English, alongside a benchmark dataset for standard English. We manually add sarcasm explanations to an existing sarcasm-labeled dataset for Australian and Indian English called BESSTIE, and compare the performance for explainable sarcasm detection for them with FLUTE, a standard English dataset containing sarcasm explanations. Our approach utilising PMP when evaluated on two open-weight LLMs (GEMMA and LLAMA) achieves statistically significant performance improvement across all tasks and datasets when compared with four alternative prompting strategies. We also find that alternative techniques such as agentic prompting mitigate context-related failures by enabling external knowledge retrieval. The focused contribution of our work is utilising PMP in generating sarcasm explanations for varieties of English.

cs.CL↗

arXiv2025

BESSTIE: A Benchmark for Sentiment and Sarcasm Classification for Varieties of English

Despite large language models (LLMs) being known to exhibit bias against non-standard language varieties, there are no known labelled datasets for sentiment analysis of English. To address this gap, we introduce BESSTIE, a benchmark for sentiment and sarcasm classification for three varieties of English: Australian (en-AU), Indian (en-IN), and British (en-UK). We collect datasets for these language varieties using two methods: location-based for Google Places reviews, and topic-based filtering for Reddit comments. To assess whether the dataset accurately represents these varieties, we conduct two validation steps: (a) manual annotation of language varieties and (b) automatic language variety prediction. Native speakers of the language varieties manually annotate the datasets with sentiment and sarcasm labels. We perform an additional annotation exercise to validate the reliance of the annotated labels. Subsequently, we fine-tune nine LLMs (representing a range of encoder/decoder and mono/multilingual models) on these datasets, and evaluate their performance on the two tasks. Our results show that the models consistently perform better on inner-circle varieties (i.e., en-AU and en-UK), in comparison with en-IN, particularly for sarcasm classification. We also report challenges in cross-variety generalisation, highlighting the need for language variety-specific datasets such as ours. BESSTIE promises to be a useful evaluative benchmark for future research in equitable LLMs, specifically in terms of language varieties. The BESSTIE dataset is publicly available at: https://huggingface.co/ datasets/unswnlporg/BESSTIE.

cs.CL↗

arXiv2025

Predicting the Target Word of Game-playing Conversations using a Low-Rank Dialect Adapter for Decoder Models

Dialect adapters that improve the performance of LLMs for NLU tasks on certain sociolects/dialects/national varieties ('dialects' for the sake of brevity) have been reported for encoder models. In this paper, we extend the idea of dialect adapters to decoder models in our architecture called LoRDD. Using MD-3, a publicly available dataset of word game-playing conversations between dialectal speakers, our task is Target Word Prediction (TWP) from a masked conversation. LoRDD combines task adapters and dialect adapters where the latter employ contrastive learning on pseudo-parallel conversations from MD-3. Our experiments on Indian English and Nigerian English conversations with two models (Mistral and Gemma) demonstrate that LoRDD outperforms four baselines on TWP. Additionally, it significantly reduces the performance gap with American English, narrowing it to 12% and 5.8% for word similarity, and 25% and 4.5% for accuracy, respectively. The focused contribution of LoRDD is in its promise for dialect adaptation of decoder models using TWP, a simplified version of the commonly used next-word prediction task.

cs.CL↗

arXiv2024

Evaluating Dialect Robustness of Language Models via Conversation Understanding

With an evergrowing number of LLMs reporting superlative performance for English, their ability to perform equitably for different dialects of English ($\textit{i.e.}$, dialect robustness) needs to be ascertained. Specifically, we use English language (US English or Indian English) conversations between humans who play the word-guessing game of 'taboo'. We formulate two evaluative tasks: target word prediction (TWP) ($\textit{i.e.}$, predict the masked target word in a conversation) and target word selection (TWS) ($\textit{i.e.}$, select the most likely masked target word in a conversation, from among a set of candidate words). Extending MD3, an existing dialectic dataset of taboo-playing conversations, we introduce M-MD3, a target-word-masked version of MD3 with the en-US and en-IN subsets. We create two subsets: en-MV (where en-US is transformed to include dialectal information) and en-TR (where dialectal information is removed from en-IN). We evaluate one open-source (Llama3) and two closed-source (GPT-4/3.5) LLMs. LLMs perform significantly better for US English than Indian English for both TWP and TWS tasks, for all settings, exhibiting marginalisation against the Indian dialect of English. While GPT-based models perform the best, the comparatively smaller models work more equitably after fine-tuning. Our error analysis shows that the LLMs can understand the dialect better after fine-tuning using dialectal data. Our evaluation methodology exhibits a novel way to examine attributes of language models using pre-existing dialogue datasets.

cs.CL↗

arXiv2024

Experiences from Creating a Benchmark for Sentiment Classification for Varieties of English

Existing benchmarks often fail to account for linguistic diversity, like language variants of English. In this paper, we share our experiences from our ongoing project of building a sentiment classification benchmark for three variants of English: Australian (en-AU), Indian (en-IN), and British (en-UK) English. Using Google Places reviews, we explore the effects of various sampling techniques based on label semantics, review length, and sentiment proportion and report performances on three fine-tuned BERT-based models. Our initial evaluation reveals significant performance variations influenced by sample characteristics, label semantics, and language variety, highlighting the need for nuanced benchmark design. We offer actionable insights for researchers to create robust benchmarks, emphasising the importance of diverse sampling, careful label definition, and comprehensive evaluation across linguistic varieties.

cs.CL↗