LiT (Lost in Translation) Robustness - Syntactic Complexity and Garden Paths: 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 garden-path and complex syntax 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: garden-path and complex syntax; 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-reasoning77.1#175
2GLM-563.5#269
3Qwen 3.5 397B A17B62.5#264
4Qwen 3 235B A22B 2507 (Thinking)51#466
5Kimi K2 (Thinking)51
6GLM-4.746.9#380
7Gemma 4 31B (IT)Thinking46.9#350
8DeepSeek V3.2 ExpNon-reasoning44.8#439
9MiniMax-M2.542.7#600
10Qwen 3.5 35B A3B37.5#492
11Kimi K231.5#443
12MiMo-V2-FlashNon-reasoning30.2#684
13GPT-OSS-120BHigh29.2#704
14Gemma 3 27B (IT)27.1#1324
15Qwen 3 235B A22B 2507 Instruct26#616

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-syntactic-complexity-and-garden-paths · How It Works · Data refreshed daily, snapshot 2026-11-21.