ShriNep@EEUCA 2026: RAKSHAK - Multi-Task DeBERTa with Rationale Distillation and Jigsaw-Augmented Training for Toxic Intent Classification

• URL: http://arxiv.org/abs/2607.20450v1
• Date: Thu, 14 May 2026 10:40:27 GMT
• Title: ShriNep@EEUCA 2026: RAKSHAK - Multi-Task DeBERTa with Rationale Distillation and Jigsaw-Augmented Training for Toxic Intent Classification
• Abstract summary: This paper presents two systems for classifying World of Tanks chat utterances into six fine-grained toxic intent categories.<n>RakSHAK is a multi-task framework combining rationale distillation from Qwen2.5-14B, Supervised Contrastive Loss, and dedicated rare-class binary heads.<n>M1 fine-tunes DeBERTa-v3-base with Focal Loss on the original GameTox data plus the same 100 extremism samples, without Jigsaw transfer.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: This paper presents two systems for the GameTox Shared Task at the Workshop on EEUCA at ACL 2026, which requires classifying World of Tanks chat utterances into six fine-grained toxic intent categories (Labels 0-5). Severe class imbalance, domain-specific multilingual slang, and extremely scarce data for rare categories such as Threats (Label 4, 60 samples) and Extremism (Label 5, 24 samples) make this a challenging classification problem. Our primary submission, RAKSHAK (rak s. aka, Sanskrit for "Protector"), is a multi-task DeBERTa-v3-base (He et al., 2022) framework combining rationale distillation from Qwen2.5-14B (An et al., 2024), Supervised Contrastive Loss, and dedicated rare-class binary heads. RAKSHAK's training data is augmented with cross-domain transfer from the Jigsaw Toxic Comment dataset (16,225 samples mapped to Labels 1-4) and 100 LLM-generated extremism samples for Label 5. Our secondary system (M1) fine-tunes DeBERTa-v3-base with Focal Loss on the original GameTox data plus the same 100 extremism samples, without Jigsaw transfer. RAKSHAK achieves a Macro F1 of 0.5883 on the official test set, ranking 7th out of 35 participating teams, while M1 achieves 0.5252 Macro F1. An ablation comparing M1 with and without Jigsaw data shows that cross-domain transfer accounts for +2.6 F1 points, while RAKSHAK's multi-task architecture contributes a further +3.7 points.

Related papers

• IROH: Insightful Ranking Of Humor using Multi-Stage Hybrid Retrieval with Rationale-Distilled LLM Judges for JOKER 2026 Track Task 1 English [1.0165430007278446]
  IROH is a three-stage retrieval system for JOKER Task 1 English at CLEF 2026.<n>Our pipeline combines hybrid sparse-dense retrieval, cross-encoder reranking, and a LoRA-adapted Large Language Model judge ensemble.
  arXiv  Detail & Related papers  (2026-09-14T14:12:38Z)
• Two-Stage Mixture-of-LoRA for Multi-Task Medical Vision-Language Learning [11.067337708907017]
  We present textbfTwo-Stage Mixture-of-LoRA, a framework built on MedGemma-1.5-4B.<n>The framework uses a shared-specific Mixture-of-LoRA architecture comprising one shared LoRA and six task-specific expert LoRAs.<n>The proposed method achieves 0.85 balanced accuracy for classification, 0.48 micro-F1 for multi-label classification, 0.79 detection F1, and 17.39 regression MAE.
  arXiv  Detail & Related papers  (2026-09-13T07:23:52Z)
• IndicQE-APE: A Benchmark for Quality Estimation and Automatic Post-Editing for Indic Languages [77.88427845181154]
  We consolidate the WMT 2020--2024 shared-task lineage with an extended English--Malayalam resource into indicqe: $126,754$ instances over nine directional pairs.<n>We benchmark six prompted LLMs and three COMET metrics on segment-level QE, and three systems on APE.
  arXiv  Detail & Related papers  (2026-08-17T09:50:52Z)
• Knowledge Index of Noah's Ark [63.143852586221534]
  We introduce KINA, an 899-item benchmark across 261 fine-grained disciplines.<n>We prove a bonus-on-bar tournament weakly FOSD-dominates flat payment in released-review quality.<n>The top model, Gemini-3.1-Pro-Preview, reaches 53.17%, followed by Claude-Opus-4.6 at 49.92% and GPT-5.4 at 48.55%.
  arXiv  Detail & Related papers  (2026-06-03T17:06:49Z)
• Step-wise Rubric Rewards for LLM Reasoning [72.17879367869503]
  Reinforcement Learning with Verifiable Rewards (RLVR) is widely used to improve reasoning in large language models.<n>We find 18.2% of steps in correct responses are wrong yet positively rewarded.<n>Across six mathematical reasoning benchmarks, SRaR improves average accuracy over RaR by 3.57 points.
  arXiv  Detail & Related papers  (2026-05-17T07:08:14Z)
• thaulab@EEUCA 2026: Who Said What to Whom? A Targeting-Aware Neural-Symbolic Pipeline for Gaming Toxicity Detection [4.187230531998879]
  This paper describes our system for the EEUCA 2026 Shared Task on toxicity classification in gaming chat.<n>Our system achieves a Macro F1 of 0.6441 and accuracy of 0.9062 on the official test set, ranking 3rd in Macro F1 and 1st in accuracy among all teams.
  arXiv  Detail & Related papers  (2026-05-13T15:15:52Z)
• PSK@EEUCA 2026: Fine-Tuning Large Language Models with Synthetic Data Augmentation for Multi-Class Toxicity Detection in Gaming Chat [0.0]
  This paper describes our system for the EEUCA 2026 Shared Task on Understanding Toxic Behavior in Gaming Communities.<n>The task involves classifying World of Tanks chat messages into six toxicity categories: Non-toxic, Insults/Flaming, Other Offensive, Hate/Harassment, Threats, and Extremism.<n>We explore multiple approaches including encoder-based models, instruction-tuned LLMs with LoRA fine-tuning, hierarchical classification, one-vs-rest strategies, and various ensemble methods.
  arXiv  Detail & Related papers  (2026-05-08T03:48:31Z)
• SG-UniBuc-NLP at SemEval-2026 Task 6: Multi-Head RoBERTa with Chunking for Long-Context Evasion Detection [0.6445605125467574]
  Our system classifies English political interview responses by coarse-grained clarity (3-way) and fine-grained evasion strategy (9-way)<n>A shared RoBERTa-large encoder supplies two task-specific heads trained jointly via a multi-task objective.<n>Our system achieves a Macro-F1 of 0.80 on Subtask 1 and 0.51 on Subtask 2, ranking 11th in both subtasks.
  arXiv  Detail & Related papers  (2026-04-29T07:37:03Z)
• The 1st Winner for 5th PVUW MeViS-Text Challenge: Strong MLLMs Meet SAM3 for Referring Video Object Segmentation [65.24213788883016]
  This report presents our winning solution to the 5th PVUW MeViS-Text Challenge.<n>We build a fully training-free pipeline that combines strong multimodal large language models with SAM3.<n>Our method ranks first on the PVUW 2026 MeViS-Text test set, achieving a Final score of 0.909064 and a J&F score of 0.7897026.
  arXiv  Detail & Related papers  (2026-04-01T02:42:30Z)
• MobileLLM-R1: Exploring the Limits of Sub-Billion Language Model Reasoners with Open Training Recipes [60.57770396565211]
  We show that strong reasoning abilities can emerge with far less data.<n>MobileLLM-R50M achieves an AIME score of 15.5, compared to just 0.6 for OLMo-2-1.48B and 0.3 for SmolLM-2-1.7B.
  arXiv  Detail & Related papers  (2025-09-29T15:43:59Z)
• ThangDLU at #SMM4H 2024: Encoder-decoder models for classifying text data on social disorders in children and adolescents [49.00494558898933]
  This paper describes our participation in Task 3 and Task 5 of the #SMM4H (Social Media Mining for Health) 2024 Workshop. Task 3 is a multi-class classification task centered on tweets discussing the impact of outdoor environments on symptoms of social anxiety. Task 5 involves a binary classification task focusing on tweets reporting medical disorders in children. We applied transfer learning from pre-trained encoder-decoder models such as BART-base and T5-small to identify the labels of a set of given tweets.
  arXiv  Detail & Related papers  (2024-04-30T17:06:20Z)
• SemEval-2024 Shared Task 6: SHROOM, a Shared-task on Hallucinations and Related Observable Overgeneration Mistakes [48.83290963506378]
  This paper presents the results of the SHROOM, a shared task focused on detecting hallucinations. We observe a number of key trends in how this approach was tackled. While a majority of the teams did outperform our proposed baseline system, the performances of top-scoring systems are still consistent with a random handling of the more challenging items.
  arXiv  Detail & Related papers  (2024-03-12T15:06:22Z)
• Attention at SemEval-2023 Task 10: Explainable Detection of Online Sexism (EDOS) [15.52876591707497]
  We have worked on interpretability, trust, and understanding of the decisions made by models in the form of classification tasks. The first task consists of determining Binary Sexism Detection. The second task describes the Category of Sexism. The third task describes a more Fine-grained Category of Sexism.
  arXiv  Detail & Related papers  (2023-04-10T14:24:52Z)
• aschern at SemEval-2020 Task 11: It Takes Three to Tango: RoBERTa, CRF, and Transfer Learning [22.90521056447551]
  We describe our system for SemEval-2020 Task 11 on Detection of Propaganda Techniques in News Articles. We developed ensemble models using RoBERTa-based neural architectures, additional CRF layers, transfer learning between the two subtasks, and advanced post-processing to handle the multi-label nature of the task.
  arXiv  Detail & Related papers  (2020-08-06T18:45:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.