Pith. sign in

REVIEW 7 cited by

Dice Loss for Data-imbalanced NLP Tasks

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1911.02855 v5 pith:3UF7DYTA submitted 2019-11-07 cs.CL

Xiaoya Li , Xiaofei Sun , Yuxian Meng , Guoyin Wang , Junjun Liang , Fei Wu , Jiwei Li This is my paper
classification cs.CL
keywords examplestrainingtasksdicelossobjectiveeasy-negativeresults
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Many NLP tasks such as tagging and machine reading comprehension are faced with the severe data imbalance issue: negative examples significantly outnumber positive examples, and the huge number of background examples (or easy-negative examples) overwhelms the training. The most commonly used cross entropy (CE) criteria is actually an accuracy-oriented objective, and thus creates a discrepancy between training and test: at training time, each training instance contributes equally to the objective function, while at test time F1 score concerns more about positive examples. In this paper, we propose to use dice loss in replacement of the standard cross-entropy objective for data-imbalanced NLP tasks. Dice loss is based on the Sorensen-Dice coefficient or Tversky index, which attaches similar importance to false positives and false negatives, and is more immune to the data-imbalance issue. To further alleviate the dominating influence from easy-negative examples in training, we propose to associate training examples with dynamically adjusted weights to deemphasize easy-negative examples.Theoretical analysis shows that this strategy narrows down the gap between the F1 score in evaluation and the dice loss in training. With the proposed training objective, we observe significant performance boost on a wide range of data imbalanced NLP tasks. Notably, we are able to achieve SOTA results on CTB5, CTB6 and UD1.4 for the part of speech tagging task; SOTA results on CoNLL03, OntoNotes5.0, MSRA and OntoNotes4.0 for the named entity recognition task; along with competitive results on the tasks of machine reading comprehension and paraphrase identification.

Discussion (0). Sign in to comment.

Forward citations

Cited by 7 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. VFAD: Variational Semantic Prompting Meets Frequency-Adaptive Representation Learning for Zero-Shot Anomaly Detection

    cs.CV 2026-07 conditional novelty 6.0 of 10

    VFAD improves zero-shot anomaly detection by injecting variational, patch-derived semantic prompts into the text encoder and refining visual features with wavelet-based frequency experts.

  2. Text Distance from Nested and Hierarchical Repetitions: A Compression-Based Perspective

    cs.CL 2026-06 conditional novelty 6.0 of 10

    Ladderpath-derived distances (NCD_lp, L_Dice, L_Jaccard) with k-NN outperform gzip-NCD and BERT on out-of-distribution and few-shot text classification without training.

  3. Motion-aware Event Suppression for Event Cameras

    cs.CV 2026-02 conditional novelty 6.0 of 10

    A real-time event-camera model jointly segments independently moving objects and forecasts their motion, enabling anticipatory suppression of dynamic events, with state-of-the-art results on EVIMO.

  4. Defect-aware Hybrid Prompt Optimization via Progressive Tuning for Zero-Shot Multi-type Anomaly Detection and Segmentation

    cs.CV 2025-12 conditional novelty 6.0 of 10

    DAPO learns shared defect-aware text prompts that let CLIP segment known and novel defect types in unseen industrial domains.

  5. Implicit Counterfactual Learning for Audio-Visual Segmentation

    cs.CV 2025-07 conditional novelty 6.0 of 10

    Implicit text features and diffusion-based counterfactual samples improve audio-visual segmentation, achieving state-of-the-art results on AVS-Object and AVS-Semantic.

  6. Temporal-Conditional Referring Video Object Segmentation with Noise-Free Text-to-Video Diffusion Model

    cs.CV 2025-08 unverdicted novelty 4.0 of 10

    Removing the noise-prediction branch from a text-to-video diffusion feature extractor, plus a temporal context mask refinement module, yields claimed state-of-the-art referring video object segmentation on four benchmarks.

  7. IADGPT: Unified LVLM for Few-Shot Industrial Anomaly Detection, Localization, and Reasoning via In-Context Learning

    cs.CV 2025-08 reject novelty 4.0 of 10

    The advertised IADGPT framework and its anomaly-detection results are absent from the supplied full text, which instead reports a RAG-based cybersecurity incident-response system.

Pith tools