arXiv subjects

Divyansh Singh

Publications and source records attributed to Divyansh Singh.

5 recordsLinked to original sources

arXiv2026

Evidence Lock Before Commitment: A Frozen Interface Degrades LLM-as-Judge Evaluation

LLM judges are often asked to extract criteria and evidence before choosing between candidate answers. This workflow assumes that the intermediate record preserves the information needed for a later verdict. For reasoning-capable models, visible field order does not reveal internal decision order, so we test an observable alternative: persist the evidence in one call and make it the exclusive input to the next. Across 24,000 judgments over HelpSteer3, FeedbackQA, and CoVal, we compare standard pairwise judging, structured one-call judging, two-call evidence locking, and three-call pointwise locking with Claude Sonnet 4.5 and GPT-5. Evidence locking reduces agreement with released human preferences by 4 to 6 percentage points and increases answer-order inconsistency by 8 to 10 points relative to structured one-call judging. Pointwise locking is also harmful, while structured evidence elicitation remains close to standard judging. The result holds for both judges and all three datasets. Persisted evidence can support auditability, but it should not replace the source answers at decision time.

cs.CL↗

arXiv2026

RADAR: Rubric-Aware Dependency and Redundancy Analysis for LLM-as-Judge Evaluation

Rubric-based LLM-as-judge pipelines often assume that evaluation criteria provide independent signals. In practice, however, criteria can be behaviorally coupled: improving one criterion may systematically change scores on another, distorting aggregate scores used in model-release or product-update decisions. We introduce RADAR, a lightweight preflight diagnostic framework for estimating such coupling before large-scale evaluation. Given a rubric, RADAR generates targeted synthetic probes, scores each probe on all criteria, and produces a directional coupling matrix that shows which criteria co-score and how. We validate RADAR on three industry-relevant evaluation settings: NVIDIA HelpSteer2, SumPubMed, and the Yale-Salesforce SummEval benchmark. Using only a small number of probes per criterion, RADAR recovers human inter-criterion correlation structure (Pearson r > 0.84) and provides practitioners with concrete audit signals about redundancy, hierarchy, and aggregation sensitivity before committing to large-scale judging.

cs.CL↗

arXiv2025

SLIDE: Sliding Localized Information for Document Extraction

Constructing accurate knowledge graphs from long texts and low-resource languages is challenging, as large language models (LLMs) experience degraded performance with longer input chunks. This problem is amplified in low-resource settings where data scarcity hinders accurate entity and relationship extraction. Contextual retrieval methods, while improving retrieval accuracy, struggle with long documents. They truncate critical information in texts exceeding maximum context lengths of LLMs, significantly limiting knowledge graph construction. We introduce SLIDE (Sliding Localized Information for Document Extraction), a chunking method that processes long documents by generating local context through overlapping windows. SLIDE ensures that essential contextual information is retained, enhancing knowledge graph extraction from documents exceeding LLM context limits. It significantly improves GraphRAG performance, achieving a 24% increase in entity extraction and a 39% improvement in relationship extraction for English. For Afrikaans, a low-resource language, SLIDE achieves a 49% increase in entity extraction and an 82% improvement in relationship extraction. Furthermore, it improves upon state-of-the-art in question-answering metrics such as comprehensiveness, diversity and empowerment, demonstrating its effectiveness in multilingual and resource-constrained settings.

cs.CL↗

arXiv2021

Detection of Emotions in Hindi-English Code Mixed Text Data

Hindi-English code-mixing, the alternation between the two languages within a single utterance, accounts for a substantial share of user-generated content in India. Because such text is written in the Latin script without a standardised transliteration convention, one underlying word surfaces in many spellings and many tokens fall outside English lexica. This paper addresses emotion detection in this setting as a four-way classification problem over anger, fear, happiness and sadness, a finer-grained task than the polarity classification targeted by most prior work. We make three contributions. First, we construct a corpus of 1,589 code-mixed sentences drawn from Twitter and from video-platform comment sections, annotated by two bilingual speakers with an inter-annotator agreement of kappa = 0.94; the video-comment portion covers a longer, less hashtag-dominated register than the Twitter-only corpora used previously. Second, we propose a normalisation procedure that clusters transliteration variants by combining distributional similarity over skip-gram vectors with a hard constraint on consonant identity, since variation is carried almost entirely by vowels. Third, we compare five baselines under a common five-fold cross-validation protocol: Naive Bayes over character and over word n-grams, a word-level LSTM, a sub-word LSTM with convolved character embeddings, and an SVM over frequency-based word vectors. The sub-word LSTM attains the highest accuracy, 76.6%, against a majority-class floor of 30.8%, and the controlled comparison against an otherwise identical word-level LSTM isolates sub-word representation as the source of that gain. On macro-averaged F1, however, it ties with word n-gram Naive Bayes at 0.77, a margin that five-fold cross-validation on 1,589 instances cannot resolve. We report a per-class error analysis and discuss the limitations imposed by corpus scale.

cs.CL↗

arXiv2021

Polyth-Net: Classification of Polythene Bags for Garbage Segregation Using Deep Learning

Polythene has always been a threat to the environment since its invention. It is non-biodegradable and very difficult to recycle. Even after many awareness campaigns and practices, Separation of polythene bags from waste has been a challenge for human civilization. The primary method of segregation deployed is manual handpicking, which causes a dangerous health hazards to the workers and is also highly inefficient due to human errors. In this paper I have designed and researched on image-based classification of polythene bags using a deep-learning model and its efficiency. This paper focuses on the architecture and statistical analysis of its performance on the data set as well as problems experienced in the classification. It also suggests a modified loss function to specifically detect polythene irrespective of its individual features. It aims to help the current environment protection endeavours and save countless lives lost to the hazards caused by current methods.

cs.CV↗