LiT (Lost in Translation) - Discourse Coherence: leaderboard
Metric: MQM>=80 rate (%): share of round-trip translations whose back-translation to English scores at least 80 on an MQM error scale (minor -1, major -5, critical -25) from a Grok 4.1 Fast judge comparing it with the English source; each passage is translated serially through a four-language sequence and back, averaged over eight sequences (high, medium and low-resource languages); pragmatics passages on referential chains and connectives; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GLM-5 | 77.4 |
| 2 | Qwen 3.5 397B A17B | 75.6 |
| 3 | Gemma 4 31B (IT) | 75 |
| 4 | Gemma 4 31B (IT) (Thinking) | 74.4 |
| 5 | Kimi K2 (Thinking) | 70.8 |
| 6 | GLM-4.7 | 70.2 |
| 7 | Qwen 3.5 397B A17B (Non-reasoning) | 69.6 |
| 8 | Qwen 3 235B A22B 2507 (Thinking) | 66.7 |
| 9 | DeepSeek V3.2 Exp (Thinking) | 61.9 |
| 10 | GLM-5 (Non-reasoning) | 60.7 |
| 11 | Qwen 3.5 35B A3B | 60.1 |
| 12 | DeepSeek V3.2 Exp (Non-reasoning) | 57.1 |
| 13 | Kimi K2 | 55.6 |
| 14 | Gemma 3 27B (IT) | 55.4 |
| 15 | GPT-OSS-120B (High) | 54.8 |
Interactive version: theaggregate.ai/benchmark?slug=lit-lost-in-translation-discourse-coherence · How It Works · Data refreshed daily, snapshot 2026-10-07.