# Open-ended Question Answering on TruthfulQA (Specific Accuracy Metrics) Source: [Open-ended Question Answering on TruthfulQA (Specific Accuracy Metrics)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:53:37.974Z Evaluation of model performance on the TruthfulQA dataset using specific accuracy breakdowns: Neutral, Positive (+), Negative (-), and Overall Accuracy. 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 | Open-ended Question Answering | | Dataset (including any stated split) | TruthfulQA | | Metric | Neutral Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 15 of 15 | | 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 - [Neutral Accuracy](): Higher is better; unit: PERCENT. - [Positive Accuracy (+)](): Higher is better; unit: PERCENT. - [Negative Accuracy (-)](): Higher is better; unit: PERCENT. - [Overall Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Neutral Accuracy | Result date | Report paper | | --- | --- | --- | --- | | MoLaCE | 74.24 | Not recorded | 69535c9a53fdd5c36173db7f | | MoLaCE + Debate | 73.74 | Not recorded | 69535c9a53fdd5c36173db7f | | Raw model | 64.22 | Not recorded | 69535c9a53fdd5c36173db7f | | Debate | 60.61 | Not recorded | 69535c9a53fdd5c36173db7f | | MoLaCE + Debate | 60.26 | Not recorded | 69535c9a53fdd5c36173db7f | | MoLaCE + Debate | 55.56 | Not recorded | 69535c9a53fdd5c36173db7f | | MoLaCE | 51.05 | Not recorded | 69535c9a53fdd5c36173db7f | | Raw model | 48.76 | Not recorded | 69535c9a53fdd5c36173db7f | | Debate+ | 46.63 | Not recorded | 69535c9a53fdd5c36173db7f | | MoLaCE | 45.11 | Not recorded | 69535c9a53fdd5c36173db7f | | Debate | 33.33 | Not recorded | 69535c9a53fdd5c36173db7f | | Debate | 30.3 | Not recorded | 69535c9a53fdd5c36173db7f | | Debate+ | 30.27 | Not recorded | 69535c9a53fdd5c36173db7f | | Raw model | 26.97 | Not recorded | 69535c9a53fdd5c36173db7f | | Debate+ | 25.09 | Not recorded | 69535c9a53fdd5c36173db7f | ### MoLaCE Neutral Accuracy: 74.24. Result ID: 6996ac025698665ec4f610f1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Mistral"}. This record does not assert independent reproduction. ### MoLaCE + Debate Neutral Accuracy: 73.74. Result ID: 6996ac025698665ec4f610f2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Mistral"}. This record does not assert independent reproduction. ### Raw model Neutral Accuracy: 64.22. Result ID: 6996ac025698665ec4f610ee. [Reporting paper]() Recorded evaluation configuration: {"Model":"Mistral"}. This record does not assert independent reproduction. ### Debate Neutral Accuracy: 60.61. Result ID: 6996ac025698665ec4f610ef. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Mistral"}. This record does not assert independent reproduction. ### MoLaCE + Debate Neutral Accuracy: 60.26. Result ID: 6996ac025698665ec4f610f7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaMA"}. This record does not assert independent reproduction. ### MoLaCE + Debate Neutral Accuracy: 55.56. Result ID: 6996ac025698665ec4f610ed. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Phi"}. This record does not assert independent reproduction. ### MoLaCE Neutral Accuracy: 51.05. Result ID: 6996ac025698665ec4f610f6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaMA"}. This record does not assert independent reproduction. ### Raw model Neutral Accuracy: 48.76. Result ID: 6996ac025698665ec4f610f3. [Reporting paper]() Recorded evaluation configuration: {"Model":"LLaMA"}. This record does not assert independent reproduction. ### Debate+ Neutral Accuracy: 46.63. Result ID: 6996ac025698665ec4f610f0. [Reporting paper]() Recorded evaluation configuration: {"Model":"Mistral"}. This record does not assert independent reproduction. ### MoLaCE Neutral Accuracy: 45.11. Result ID: 6996ac025698665ec4f610ec. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Phi"}. This record does not assert independent reproduction. ### Debate Neutral Accuracy: 33.33. Result ID: 6996ac025698665ec4f610f4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLaMA"}. This record does not assert independent reproduction. ### Debate Neutral Accuracy: 30.3. Result ID: 6996ac025698665ec4f610ea. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Phi"}. This record does not assert independent reproduction. ### Debate+ Neutral Accuracy: 30.27. Result ID: 6996ac025698665ec4f610f5. [Reporting paper]() Recorded evaluation configuration: {"Model":"LLaMA"}. This record does not assert independent reproduction. ### Raw model Neutral Accuracy: 26.97. Result ID: 6996ac025698665ec4f610e9. [Reporting paper]() Recorded evaluation configuration: {"Model":"Phi"}. This record does not assert independent reproduction. ### Debate+ Neutral Accuracy: 25.09. Result ID: 6996ac025698665ec4f610eb. [Reporting paper]() Recorded evaluation configuration: {"Model":"Phi"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.