Qwen Chinese human evaluation
Qwen-14B 30.7% Three annotators rank chat model answers to 300 Chinese instructions on knowledge, language, writing, math and coding. Results are win rates against GPT-3.5.
In-house Qwen evaluation. The instructions are not public.
- Made by
- Qwen Team, Alibaba Group
- Released
- September 2023
- Kind
- Overall chat quality
- Level
- Everyday
- Scored
- Ranked by votes
- Answers
- Open-ended
- Tasks
- Kept private
- Strongest model tested
- Qwen-14B #213 on the capability index
- Last in a system card
- September 2023
- Last model added
- September 2023
2 models have ratings here, all from their labs’ own posts
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs. A rating ranks a model only against others rated in the same arena at the same time, so these don’t show who leads.
Source: Labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Labs’ model cards
Labs report it only in system cards
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark. Some system cards do, below.
2 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Sep 28, 2023 | Qwen-7B technical report | 28.4 | Human evaluation on 300 Chinese instructions, average win rate vs GPT-3.5 for the SFT chat model, tie 28.3, loss 43.3, printed on the chart (Figure 4). | Capabilities |
| Sep 28, 2023 | Qwen-14B technical report | 30.7 | Human evaluation on 300 Chinese instructions, average win rate vs GPT-3.5 for the SFT chat model, tie 32.3, loss 37.0, printed on the chart (Figure 4). RLHF version: win 38.7, tie 28.9, loss 32.4 | Capabilities |
More in Instructions and chat
- IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy.
Gemma 4 31B IT 98.9% - IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose).
Nemotron-Cascade 2 30B-A3B 82.9% - MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
- Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
- AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics.
MAI-Thinking-1 85% - IFStruct In use Models generate JSON or YAML that must follow a requested schema and phrasing constraints, without constrained decoding. A response passes only if every structural constraint holds. Gemma 4 31B IT 95.9%