LiT (Lost in Translation) Robustness - Conceptual and Abstract Nuance: leaderboard

LiT, Lost in Translation (2026), measures multilingual proficiency by round-trip translation: English passages pass through a sequence of four languages and back, and an LLM judge scores the result against the source with MQM error weights. Share of robustness sentences with metaphorical use of physical words whose round trips score at least 80, averaged over eight translation sequences, in percent; higher is better.

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); robustness subset: metaphorical use of physical vocabulary; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 17 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 FlashNon-reasoning83.3#175
2Qwen 3.5 397B A17B75#264
3Gemma 4 31B (IT)Thinking75#350
4GLM-574#269
5Kimi K2 (Thinking)67.7
6GLM-4.761.5#380
7Qwen 3 235B A22B 2507 (Thinking)60.4#466
8Gemma 3 27B (IT)57.3#1324
9Qwen 3.5 35B A3B57.3#492
10DeepSeek V3.2 ExpThinking57.3#439
11Kimi K251.4#443
12MiniMax-M2.543.8#600
13Qwen 3 235B A22B 2507 Instruct41.7#616
14MiMo-V2-FlashNon-reasoning39.6#684
15GPT-OSS-120BHigh37.5#704

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=lit-lost-in-translation-robustness-conceptual-and-abstract-nuance · How It Works · Data refreshed daily, snapshot 2026-11-21.