# Fake News Stance Detection on FNC 1 (test) Source: [Fake News Stance Detection on FNC 1 (test)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:34:46.325Z Evaluation of Fake News Stance Detection performance on the test split of the FNC 1 dataset, using metrics such as Score\_FNC, Agree, Disagree, Discuss, Unrelated, and Overall. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Fake News Stance Detection | | Dataset (including any stated split) | FNC 1 (test) | | Metric | Agree | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 32 of 32 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Agree](): Higher is better; unit: PERCENT. - [Score\_FNC](): Higher is better; unit: PERCENT. - [Disagree](): Higher is better; unit: PERCENT. - [Discuss](): Higher is better; unit: PERCENT. - [Unrelated](): Higher is better; unit: PERCENT. - [Overall](): Higher is better; unit: PERCENT. - [Weighted Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Agree | Result date | Report paper | | --- | --- | --- | --- | | CNN + biLSTM | 74.09 | Not recorded | 68fa63d71e209131cea01255 | | Stacked bi-LSTMs + shortcuts + max-pooling + attention | 66.47 | Not recorded | 68fa69141e209131cea2012f | | biLSTM + Attention | 58.74 | Not recorded | 68fa63d71e209131cea01255 | | SOLAT in the SWEN | 58.5 | Not recorded | 68fa63d71e209131cea01255 | | Team SOLAT in the SWEN | 58.5 | Not recorded | 68fa69141e209131cea2012f | | Stacked bi-LSTMs + shortcuts + max-pooling | 56.59 | Not recorded | 68fa69141e209131cea2012f | | Chips Ahoy! | 55.96 | Not recorded | 68fa63d71e209131cea01255 | | Stacked bi-GRUs + shortcuts + max-pooling + attention | 52.86 | Not recorded | 68fa69141e209131cea2012f | | MLP with the considered external features | 52.55 | Not recorded | 68fa69141e209131cea2012f | | Bi-GRU + max-pooling + attention | 52.08 | Not recorded | 68fa69141e209131cea2012f | | Bi-LSTM + max-pooling + attention (best encoder) | 51.34 | Not recorded | 68fa69141e209131cea2012f | | Word2vec + External Features | 50.7 | Not recorded | 68fa63d71e209131cea01255 | | Baseline based on word2vec + hand-crafted features | 50.7 | Not recorded | 68fa69141e209131cea2012f | | Bi-LSTM + max-pooling | 45.35 | Not recorded | 68fa69141e209131cea2012f | | Athene | 44.72 | Not recorded | 68fa63d71e209131cea01255 | | Team Athene | 44.72 | Not recorded | 68fa69141e209131cea2012f | | TF-IDF baseline | 44.04 | Not recorded | 68fa63d71e209131cea01255 | | UCL Machine Reading | 44.04 | Not recorded | 68fa63d71e209131cea01255 | | Baseline based on TF-IDF vectors | 44.04 | Not recorded | 68fa69141e209131cea2012f | | Team UCL Machine Reading | 44.04 | Not recorded | 68fa69141e209131cea2012f | | Proposed | 43.82 | Not recorded | 68fa63d71e209131cea01255 | | Previous state-of-the-art – Bhatt et al. | 43.82 | Not recorded | 68fa69141e209131cea2012f | | biLSTM | 38.04 | Not recorded | 68fa63d71e209131cea01255 | | Neural baseline based on bi-directional LSTMs | 38.04 | Not recorded | 68fa69141e209131cea2012f | | CNN | 35.89 | Not recorded | 68fa63d71e209131cea01255 | | Skip-thought baseline | 31.8 | Not recorded | 68fa63d71e209131cea01255 | | Baseline based on skip-thought embeddings | 31.8 | Not recorded | 68fa69141e209131cea2012f | | FNC-1 baseline | 9.09 | Not recorded | 68fa63d71e209131cea01255 | | FNC-1 baseline | 9.09 | Not recorded | 68fa69141e209131cea2012f | | Best set of features from Tosik et al. | 1.42 | Not recorded | 68fa69141e209131cea2012f | ### CNN + biLSTM Agree: 74.09. Result ID: 698c04f4f729bc7f68e7d69d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Architecture":"CNN and biLSTM hybrid"}. This record does not assert independent reproduction. ### Stacked bi-LSTMs + shortcuts + max-pooling + attention Agree: 66.47. Result ID: 698d97ee8b517eb433228ff5. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"architecture":"Stacked bi-LSTMs","shortcuts":"true","pooling":"max-pooling","attention":"true"}. This record does not assert independent reproduction. ### biLSTM + Attention Agree: 58.74. Result ID: 698c04f4f729bc7f68e7d69c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Architecture":"Bidirectional LSTM with Attention"}. This record does not assert independent reproduction. ### SOLAT in the SWEN Agree: 58.5. Result ID: 698c04f4f729bc7f68e7d696. [Reporting paper]() Recorded evaluation configuration: {"Category":"Top-4 submission"}. This record does not assert independent reproduction. ### Team SOLAT in the SWEN Agree: 58.5. Result ID: 698d97ee8b517eb433228fed. [Reporting paper]() Recorded evaluation configuration: {"rank":"1st place at FNC-1"}. This record does not assert independent reproduction. ### Stacked bi-LSTMs + shortcuts + max-pooling Agree: 56.59. Result ID: 698d97ee8b517eb433228ff3. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"architecture":"Stacked bi-LSTMs","shortcuts":"true","pooling":"max-pooling"}. This record does not assert independent reproduction. ### Chips Ahoy! Agree: 55.96. Result ID: 698c04f4f729bc7f68e7d699. [Reporting paper]() Recorded evaluation configuration: {"Category":"Top-4 submission"}. This record does not assert independent reproduction. ### Stacked bi-GRUs + shortcuts + max-pooling + attention Agree: 52.86. Result ID: 698d97ee8b517eb433228ff4. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"architecture":"Stacked bi-GRUs","shortcuts":"true","pooling":"max-pooling","attention":"true"}. This record does not assert independent reproduction. ### MLP with the considered external features Agree: 52.55. Result ID: 698d97ee8b517eb433228fef. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"architecture":"MLP","features":"external"}. This record does not assert independent reproduction. ### Bi-GRU + max-pooling + attention Agree: 52.08. Result ID: 698d97ee8b517eb433228ff1. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"architecture":"Bi-GRU","pooling":"max-pooling","attention":"true"}. This record does not assert independent reproduction. ### Bi-LSTM + max-pooling + attention (best encoder) Agree: 51.34. Result ID: 698d97ee8b517eb433228ff2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"architecture":"Bi-LSTM","pooling":"max-pooling","attention":"true","selection":"best encoder"}. This record does not assert independent reproduction. ### Word2vec + External Features Agree: 50.7. Result ID: 698c04f4f729bc7f68e7d693. [Reporting paper]() Recorded evaluation configuration: {"Features":"Word2vec, External"}. This record does not assert independent reproduction. ### Baseline based on word2vec + hand-crafted features Agree: 50.7. Result ID: 698d97ee8b517eb433228fe4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"features":"word2vec + hand-crafted features"}. This record does not assert independent reproduction. ### Bi-LSTM + max-pooling Agree: 45.35. Result ID: 698d97ee8b517eb433228ff0. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"architecture":"Bi-LSTM","pooling":"max-pooling"}. This record does not assert independent reproduction. ### Athene Agree: 44.72. Result ID: 698c04f4f729bc7f68e7d697. [Reporting paper]() Recorded evaluation configuration: {"Category":"Top-4 submission"}. This record does not assert independent reproduction. ### Team Athene Agree: 44.72. Result ID: 698d97ee8b517eb433228fec. [Reporting paper]() Recorded evaluation configuration: {"rank":"2nd place at FNC-1"}. This record does not assert independent reproduction. ### TF-IDF baseline Agree: 44.04. Result ID: 698c04f4f729bc7f68e7d695. [Reporting paper]() Recorded evaluation configuration: {"Type":"Baseline"}. This record does not assert independent reproduction. ### UCL Machine Reading Agree: 44.04. Result ID: 698c04f4f729bc7f68e7d698. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Category":"Top-4 submission"}. This record does not assert independent reproduction. ### Baseline based on TF-IDF vectors Agree: 44.04. Result ID: 698d97ee8b517eb433228fe6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"features":"TF-IDF vectors"}. This record does not assert independent reproduction. ### Team UCL Machine Reading Agree: 44.04. Result ID: 698d97ee8b517eb433228feb. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"rank":"3rd place at FNC-1"}. This record does not assert independent reproduction. ### Proposed Agree: 43.82. Result ID: 698c04f4f729bc7f68e7d69e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Features":"Neural, Statistical and External features"}. This record does not assert independent reproduction. ### Previous state-of-the-art – Bhatt et al. Agree: 43.82. Result ID: 698d97ee8b517eb433228fee. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### biLSTM Agree: 38.04. Result ID: 698c04f4f729bc7f68e7d69b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Architecture":"Bidirectional LSTM"}. This record does not assert independent reproduction. ### Neural baseline based on bi-directional LSTMs Agree: 38.04. Result ID: 698d97ee8b517eb433228fe8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"architecture":"bi-directional LSTMs"}. This record does not assert independent reproduction. ### CNN Agree: 35.89. Result ID: 698c04f4f729bc7f68e7d69a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Architecture":"Convolutional Neural Network"}. This record does not assert independent reproduction. ### Skip-thought baseline Agree: 31.8. Result ID: 698c04f4f729bc7f68e7d694. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Type":"Baseline"}. This record does not assert independent reproduction. ### Baseline based on skip-thought embeddings Agree: 31.8. Result ID: 698d97ee8b517eb433228fe5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"features":"skip-thought embeddings"}. This record does not assert independent reproduction. ### FNC-1 baseline Agree: 9.09. Result ID: 698c04f4f729bc7f68e7d692. [Reporting paper]() Recorded evaluation configuration: {"Type":"Baseline"}. This record does not assert independent reproduction. ### FNC-1 baseline Agree: 9.09. Result ID: 698d97ee8b517eb433228fe3. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Best set of features from Tosik et al. Agree: 1.42. Result ID: 698d97ee8b517eb433228fe7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.