Vol. 1 No. 4 (2026)

Articles

Contradiction-Sensitive Routing of Evidence during Abstention Training

Published 2026-09-07

  • William Nyström
    • ,
  • Amanda Ek

William Nyström
Department of Electrical Engineering, Faculty of Science and Technology, Uppsala University, Uppsala, Sweden

Amanda Ek
Department of Electrical Engineering, Faculty of Science and Technology, Uppsala University, Uppsala, Sweden

Data Science and Computational Intelligence, Volume 1, Issue 4, 2026 cover

Abstract

The integration of multiple evidence sources in modern artificial intelligence systems frequently encounters a critical bottleneck when the retrieved information contains contradictions. Traditional neural architectures typically force a prediction even in the presence of highly conflicting inputs, leading to overconfident and erroneous outputs. Abstention training has emerged as a promising paradigm to mitigate this by allowing models to defer or reject answering when uncertainty is high. However, existing abstention mechanisms generally rely on global uncertainty metrics and fail to dynamically analyze the specific contradictions between localized pieces of evidence. This paper introduces a novel framework for contradiction-sensitive routing of evidence during abstention training. By explicitly modeling the relational conflicts between evidence pairs, the proposed architecture routes contradictory signals through specialized neural pathways that directly modulate the abstention gate. Through comprehensive theoretical modeling and empirical evaluation, this study demonstrates that contradiction-sensitive routing significantly improves the reliability of the system. The model learns to distinguish between benign ambiguity and fundamental factual conflicts, optimizing the trade-off between answer coverage and accuracy. Extensive experiments across varied datasets indicate that incorporating contradiction-aware routing mechanisms enhances the effective reliability of the model while maintaining competitive predictive performance on non-contradictory queries.

Keywords

Contradiction Sensitivity, Abstention Training, Evidence Routing, Machine Learning

PDF

References

  1. 1. Zhao, R., Tang, J., Zeng, W., Chen, Z., & Zhao, X. (2024, October). Zero-shot knowledge graph question generation via multi-agent llms and small models synthesis. In Proceedings of the 33rd ACM International Conference on Information and Knowledge Management (pp. 3341-3351).
  2. 2. Huang, Y., Thede, L., Mancini, M., Xu, W., & Akata, Z. (2025, September). Investigating structural pruning and recovery techniques for compressing multimodal large language models: An empirical study. In DAGM german conference on pattern recognition (pp. 320-336). Cham: Springer Nature Switzerland.
  3. 3. Mo, Z. A Study on Chinese-English Bilingual Question Answering Systems Using Cross-Lingual Pre-Trained Models. October 2025. Image Processing, Electronics and Computers, DOI, 10.
  4. 4. Yu, R., Wang, Y., Jiao, X., Zhang, Y., & Kwok, J. T. (2024). Direct alignment of language models via quality-aware self-refinement. arXiv preprint arXiv:2405.21040.
  5. 5. Sang, Y. (2025, July). Towards explainable rag: Interpreting the influence of retrieved passages on generation. In 2025 4th International Conference on Robotics, Artificial Intelligence and Intelligent Control (RAIIC) (pp. 397-400). IEEE.
  6. 6. Sang, Y. (2025, October). AutoCrit: A Meta-Reasoning Framework for Self-Critique and Iterative Error Correction in LLM Chains-of-Thought. In 2025 6th International Conference on Machine Learning and Computer Application (ICMLCA) (pp. 1177-1180). IEEE.
  7. 7. Tang, J., Wang, Z., Gong, Z., Yu, J., Zhu, X., & Yin, J. (2025, April). Multi-grained query-guided set prediction network for grounded multimodal named entity recognition. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 39, No. 24, pp. 25246-25254).
  8. 8. Tang, J., Yang, Y., Yu, J., Wang, Z. X., Liang, H., Yao, L., & Yin, J. (2025, November). Unco: Uncertainty-driven collaborative framework of large and small models for grounded multimodal ner. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (pp. 7644-7662).
  9. 9. Zhang, Y., Zhao, M., Zhang, Y., & Cheung, Y. M. (2025). Trending applications of large language models: A user perspective survey. IEEE Transactions on Artificial Intelligence.
  10. 10. Su, Yiyun, et al. "Agentic-SQL Taxonomy: A Survey of Autonomous and Interactive Text-to-SQL with LLMs." (2026).
  11. 11. Jiang, J., Yang, P., Zhang, R., & Liu, F. (2026, July). Towards efficient large language model serving: A survey on system-aware kv cache optimization. In Findings of the Association for Computational Linguistics: ACL 2026 (pp. 38450-38476).
  12. 12. Cao, Y., Jing, L., Wang, Y., Shi, D., Yu, C., & Xing, J. (2026, June). Dual-Pathway Diffusion for Hand Correction in Synthetic Portraits: Global Context Aware and Local Structure Refinement. In Proceedings of the 2026 International Conference on Multimedia Retrieval (pp. 1899-1907).
  13. 13. Zhu, D., Xie, C., Zhang, H., Wei, Z., Wang, Z., Shi, J., ... & Xie, Q. (2026). Standalone LLM and a Pre-specified Agentic Pipeline for Explaining ICU Mortality Predictions: a Feasibility Study on the eICU Demo Dataset. arXiv preprint arXiv:2608.26109.
  14. 14. Qiu, Z., Lyu, H., Xiong, W., & Luo, J. (2025). Can llms simulate social media engagement? a study on action-guided response generation. arXiv preprint arXiv:2502.12073.
  15. 15. Emani, M., Foreman, S., Sastry, V., Xie, Z., Raskar, S., Arnold, W., ... & Papka, M. E. (2023). A comprehensive performance study of large language models on novel ai accelerators. arXiv preprint arXiv:2310.04607.
  16. 16. Wang, H., Xu, Q., Liu, C., Wu, J., Lin, F., & Chen, W. (2026, April). Emergent hierarchical reasoning in llms through reinforcement learning. In International Conference on Learning Representations (Vol. 2026, pp. 74519-74543).
  17. 17. Fan, D., Liu, M., Shao, Y., Yang, L., Liu, Y., Zhang, Y., ... & Wang, Z. (2025). Domain-specific large language model for maintenance decision-making on wind farms by labeled-data-supervised fine-tuning. Engineering.
  18. 18. Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019). RoBERTa: A robustly optimized BERT pretraining approach. arXiv preprint arXiv:1907.11692.
  19. 19. Sutskever, I., Vinyals, O., & Le, Q. V. (2014). Sequence to sequence learning with neural networks. In Advances in Neural Information Processing Systems.
  20. 20. Beltagy, I., Peters, M. E., & Cohan, A. (2020). Longformer: The long-document transformer. arXiv preprint arXiv:2004.05150.
  21. 21. Banerjee, S., & Lavie, A. (2005). METEOR: An automatic metric for MT evaluation with improved correlation with human judgments. In Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for MT and/or Summarization.
  22. 22. Li, Q., Ye, Q., Zhang, N., Zhang, W., & Hu, F. (2025). Digital-twin-enabled industrial IoT: Vision, framework, and future directions. IEEE Wireless Communications, 32(6), 173-181.
  23. 23. Li, Y. Z., Zhang, J. P., Chen, Z. Y., Wang, H. C., Zhang, K. X., Hu, S. S., ... & Dudley, M. (2026, September). Growth and Characterization of High-Quality Thick Epitaxial 4H-SiC Wafers for High Voltage Devices. In Defect and Diffusion Forum (Vol. 452, pp. 79-85). Trans Tech Publications Ltd.
  24. 24. Wang, S., Zhang, X., Wang, P., & Li, X. (2026). Design of Magnetic Sensor Array-Based PUFs for IoT Security. IEEE Transactions on Instrumentation and Measurement, 75, 1-9.
  25. 25. Li, X., Wang, P., Li, G., Ni, L., & Zhang, Y. (2023). Design of interface circuits and lightweight PUF for TMR sensors. IEEE Sensors Journal, 23(11), 11754-11761.
  26. 26. Liang, R., Xu, S., Xie, C., Chen, J., Ren, F., Yang, S., & Yabe, T. (2026). Abstain Mask Retain Core: Time Series Prediction by Adaptive Masking Loss with Representation Consistency. Advances in Neural Information Processing Systems, 38, 99445-99471.
  27. 27. Li, S. (2024). Machine Learning in Credit Risk Forecasting — — A Survey on Credit Risk Exposure. Accounting and Finance Research, 13(2), 107-107.
  28. 28. Zhao, J., Zhang, C., Qin, M., & Yang, P. (2025). QuantFactor REINFORCE: mining steady formulaic alpha factors with variance-bounded REINFORCE. IEEE Transactions on Signal Processing, 73, 2448-2463.
  29. 29. Qu, W., Wang, J., Gong, Y., Huang, X., & Xiao, L. (2025, June). An end-to-end robust point cloud semantic segmentation network with single-step conditional diffusion models. In 2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 27325-27335). IEEE.
  30. 30. Zhao, J., Zhang, C., Wang, C., & Yang, P. (2025). Learning from expert factors: Trajectory-level reward shaping for formulaic alpha mining. arXiv preprint arXiv:2507.20263.
  31. 31. Peng, Y., Li, H., BouDagher-Fadel, M., Wang, L., Zhang, D., Zheng, T., & Yang, K. (2022). Benthic foraminifera distribution and sedimentary environmental evolution of a carbonate platform: A case study of the Guadalupian (middle Permian) in eastern Sichuan Basin. Marine Micropaleontology, 170, 102079.
  32. 32. Yu, A., Huang, Y., & Xia, L. (2023). Efficient characterization of phase modulator based on Lyot-Sagnac interferometer. Measurement, 210, 112538.
  33. 33. Jia, Y., Ye, Y., Ma, Z., & Wang, T. (2024). The effect of subnational legal effectiveness and social trust on foreign firm performance: from subnational analysis in emerging economies. International Journal of Emerging Markets, 19(6), 1669-1694.
  34. 34. Wang, Y., & Ling, C. (2025). Comparing SAS® and R Approaches in Reshaping data. In PharmaSUG 2025 Conference Proceedings.
  35. 35. Grover, L. K. (1996). A fast quantum mechanical algorithm for database search. In Proceedings of STOC.