LiT (Lost in Translation) - Core Semantics: 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 pragmatics passages on truth conditions and entailment whose round trip scores at least 80, 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); pragmatics passages on truth conditions and entailment; 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-reasoning87.5#180
2Qwen 3.5 397B A17B81.2#270
3Gemma 4 31B (IT)78.1#354
4GLM-577.1#276
5GLM-4.772.4#383
6Kimi K2 (Thinking)70.6
7Qwen 3 235B A22B 2507 (Thinking)67.2#471
8DeepSeek V3.2 ExpThinking65.6#440
9Qwen 3.5 35B A3B63.5#494
10Kimi K260.9#448
11Gemma 3 27B (IT)60.4#1341
12MiniMax-M2.557.8#600
13GPT-OSS-120BHigh55.7#712
14Qwen 3 235B A22B 2507 Instruct53.6#625
15MiMo-V2-FlashNon-reasoning44.3#691

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=lit-lost-in-translation-core-semantics · How It Works · Data refreshed daily, snapshot 2026-11-22.