# Multi-hop Reasoning on RULER QA Source: [Multi-hop Reasoning on RULER QA]() Collection index: [llms.txt]() Data updated: 2026-03-04T17:01:02.911Z Evaluation of model performance on the RULER QA dataset for multi-hop reasoning, specifically reporting results across different context lengths (32K, 64K, 128K). 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 | Multi-hop reasoning | | Dataset (including any stated split) | RULER QA | | Metric | Accuracy (32K Context) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | 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 - [Accuracy (32K Context)](): Higher is better; unit: PERCENT. - [Accuracy (64K Context)](): Higher is better; unit: PERCENT. - [Accuracy (128K Context)](): Higher is better; unit: PERCENT. - [Average Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy (32K Context) | Result date | Report paper | | --- | --- | --- | --- | | Qwen2.5-14B-1M-LongRLVR | 95.4 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen3-14B (Thinking) | 91.2 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-14B-1M | 90.6 | Not recorded | 69a66c582d603e6de3e3d176 | | QwenLong-L1-32B | 89 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-14B-1M-SFT | 88 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen3-8B (Thinking) | 86.5 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-14B-1M-RLVR | 86.3 | Not recorded | 69a66c582d603e6de3e3d176 | | LLaMA-3.1-8B-LongRLVR | 85.5 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-7B-1M-LongRLVR | 82.5 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-7B-1M-RLVR | 74.4 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-7B-1M-SFT | 72.4 | Not recorded | 69a66c582d603e6de3e3d176 | | LLaMA-3.1-8B-RLVR | 72 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-7B-1M | 70.5 | Not recorded | 69a66c582d603e6de3e3d176 | | LLaMA-3.1-70B | 70.4 | Not recorded | 69a66c582d603e6de3e3d176 | | LLaMA-3.1-8B-SFT | 68.4 | Not recorded | 69a66c582d603e6de3e3d176 | | Qwen2.5-72B-YaRN | 66.9 | Not recorded | 69a66c582d603e6de3e3d176 | | LLaMA-3.1-8B | 65.8 | Not recorded | 69a66c582d603e6de3e3d176 | ### Qwen2.5-14B-1M-LongRLVR Accuracy (32K Context): 95.4. Result ID: 69a865406dbb7075b95017e4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-14B-1M","Protocol":"LongRLVR"}. This record does not assert independent reproduction. ### Qwen3-14B (Thinking) Accuracy (32K Context): 91.2. Result ID: 69a865406dbb7075b95017d7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-14B","Thinking Mode":"true"}. This record does not assert independent reproduction. ### Qwen2.5-14B-1M Accuracy (32K Context): 90.6. Result ID: 69a865406dbb7075b95017e1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-14B-1M"}. This record does not assert independent reproduction. ### QwenLong-L1-32B Accuracy (32K Context): 89. Result ID: 69a865406dbb7075b95017d8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"QwenLong-L1-32B"}. This record does not assert independent reproduction. ### Qwen2.5-14B-1M-SFT Accuracy (32K Context): 88. Result ID: 69a865406dbb7075b95017e2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-14B-1M","Protocol":"SFT"}. This record does not assert independent reproduction. ### Qwen3-8B (Thinking) Accuracy (32K Context): 86.5. Result ID: 69a865406dbb7075b95017d6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Thinking Mode":"true"}. This record does not assert independent reproduction. ### Qwen2.5-14B-1M-RLVR Accuracy (32K Context): 86.3. Result ID: 69a865406dbb7075b95017e3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-14B-1M","Protocol":"RLVR"}. This record does not assert independent reproduction. ### LLaMA-3.1-8B-LongRLVR Accuracy (32K Context): 85.5. Result ID: 69a865406dbb7075b95017dc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B","Protocol":"LongRLVR"}. This record does not assert independent reproduction. ### Qwen2.5-7B-1M-LongRLVR Accuracy (32K Context): 82.5. Result ID: 69a865406dbb7075b95017e0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-1M","Protocol":"LongRLVR"}. This record does not assert independent reproduction. ### Qwen2.5-7B-1M-RLVR Accuracy (32K Context): 74.4. Result ID: 69a865406dbb7075b95017df. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-1M","Protocol":"RLVR"}. This record does not assert independent reproduction. ### Qwen2.5-7B-1M-SFT Accuracy (32K Context): 72.4. Result ID: 69a865406dbb7075b95017de. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-1M","Protocol":"SFT"}. This record does not assert independent reproduction. ### LLaMA-3.1-8B-RLVR Accuracy (32K Context): 72. Result ID: 69a865406dbb7075b95017db. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B","Protocol":"RLVR"}. This record does not assert independent reproduction. ### Qwen2.5-7B-1M Accuracy (32K Context): 70.5. Result ID: 69a865406dbb7075b95017dd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-7B-1M"}. This record does not assert independent reproduction. ### LLaMA-3.1-70B Accuracy (32K Context): 70.4. Result ID: 69a865406dbb7075b95017d4. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-70B"}. This record does not assert independent reproduction. ### LLaMA-3.1-8B-SFT Accuracy (32K Context): 68.4. Result ID: 69a865406dbb7075b95017da. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B","Protocol":"SFT"}. This record does not assert independent reproduction. ### Qwen2.5-72B-YaRN Accuracy (32K Context): 66.9. Result ID: 69a865406dbb7075b95017d5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-72B","Extension":"YaRN"}. This record does not assert independent reproduction. ### LLaMA-3.1-8B Accuracy (32K Context): 65.8. Result ID: 69a865406dbb7075b95017d9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaMA-3.1-8B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.