๐ Multilingual & Translation¶
๐ฌ EMNLP2026 ยท 24 paper notes
๐ Same area in other venues: ๐ฌ ICLR2026 (8) ยท ๐ฌ ACL2026 (63) ยท ๐งช ICML2026 (3) ยท ๐ค AAAI2026 (9) ยท ๐ง NeurIPS2025 (11) ยท ๐น ICCV2025 (1)
๐ฅ Top topics: Translation ร8 ยท Alignment/RLHF ร3
- ALEE: Any-Language Evaluation of Embeddings via English-Centric Minimal Pairs
-
By applying controlled semantic perturbations on the English side of parallel corpora via Abstract Meaning Representation (AMR) with bidirectional NLI filtering, ALEE establishes a dynamic minimal-pair diagnostic benchmark across 275+ languages, exposing systematic cross-lingual alignment blind spots in structural and lexical semantics.
- Assessing the Impact of Language Disparity on Multilingual Linguistic Ability in Large Language Models
-
Evaluating six LLM families across 101 languages on the MultiBLiMP syntactic minimal-pair benchmark under four paradigms reveals that post-training imposes a linguistic alignment tax borne disproportionately by low-resource languages, while native-language prompting crucially recovers latent grammatical competence masked by unprompted probabilities.
- Bridging the Consistency Gap: A Noun Bank Augmented Framework for Low-Cost Long Text Translation
-
To resolve the severe tension between document-level lexical consistency and inference cost in long-text machine translation, NoBAT decouples translation into coarse serial Noun Bank construction with historical disambiguation and independent parallel sentence decoding via local noun constraint retrieval, improving entity consistency while slashing total token consumption by up to 6.3ร and achieving an 8.8ร wall-clock speedup.
- Cross-Attention Eigenvalues Know the Word Order in Multilingual Machine Translation
-
By training linear probes on the normalized eigenvalue spectra of cross-attention Gram matrices in NLLB-200, this paper demonstrates that cross-lingual word-order typology (SVO vs. SOV) is near-perfectly decodable and broadly distributed across layers and heads when translating from English into diverse target languages, yet degrades sharply when translating into fixed English, proving that cross-attention geometry is fundamentally shaped by target-side generation demands rather than source-side structure.
- Cross-lingual Biography Enrichment via Claim Extraction and Alignment
-
Addressing the pronounced knowledge asymmetry where non-English Wikipedia articles document long-tail figures with far richer detail than English Wikipedia, this paper presents a claim-based cross-lingual biography enrichment framework that extracts atomic claims into a shared English space, filters already covered information via fine-grained semantic alignment, and rewrites the English article with selected novel claims, achieving significantly higher supported additions with substantially lower hallucination rates.
- DeepSEA: A Multilingual Evaluation Benchmark for Southeast Asia
-
Google Research introduces DEEPSEA, the first comprehensive human-translated benchmark covering 12 Southeast Asian languages across high- and low-resource tiers on 5 NLP tasks, uncovering severe cross-lingual performance collapse in open-weights models and an overwhelming reliance on English cognitive representations.
- Discourse Dependency: A Continuous Criterion for Translation Difficulty
-
The paper reframes machine translation difficulty from traditional output-based quality estimates into a continuous, metric-free source-side criterion called Discourse Dependency (DDP) based on referential reach, demonstrating that frontier LLMs fail to exploit long-range document context and that standard automatic metrics suffer from systemic blindness on context-demanding segments.
- FuzhouBench: A Benchmark for Eastern Min Chinese with Tone Sandhi Annotations
-
This paper introduces FuzhouBench, the first public NLP benchmark for Eastern Min Chinese (Fuzhou dialect) comprising 22.8k dictionary entries with 11.3k tone sandhi annotations and 3.06 hours of paired audio; empirical evaluations demonstrate that frontier LLMs fail on Yngping G2P and symbolic sandhi rules, while colloquial search queries trigger severe BM25 collapse that requires a two-stage hybrid and cross-encoder retrieval pipeline to recover.
- IndicParam: Benchmark to evaluate LLMs on low-resource Indic Languages
-
IndicParam presents a human-verified graduate-level benchmark of 13,207 questions across 11 low- and extremely low-resource Indic languages derived from the official UGC-NET national examinations, systematically decoupling pure linguistic proficiency from general knowledge and demonstrating that even the top-performing model (Gemini-2.5) achieves only 58.0% average accuracy.
- Length-Adaptive Decoding for Masked Diffusion Machine Translation
-
Addressing the structural defect in fixed-canvas masked diffusion machine translation where target canvas length must be pre-allocated before denoising (causing truncation or hallucination), this paper introduces Entropy-Valley (EV), a training-free test-time length selector that scores candidate canvas lengths by mean predictive entropy across all-mask forward passes, closing 33.0%โ65.3% of the COMET-22 gap to reference lengths across EnโZh and EnโDe while matching or exceeding matched-data autoregressive LLaMA-3-8B.
- Loong: A Human-Like Long Document Translation Agent with Observe-and-Act Adaptive Context Selection
-
LOONG proposes a human-like long document translation agent that maintains a multi-granularity 3E (Essence-Exemplar-Entity) memory bank, executes a 3-step sequential observe-and-act reasoning process optimized via SFT+DPO reinforcement learning to adaptively filter contextual noise, and employs a recursive alignment-enforced algorithm, achieving up to 13.0 points average improvement over strong baselines across multiple language pairs and backbones.
- Mimir: Large-Scale Multilingual Concept Modeling
-
Mimir is the first large-scale multilingual Large Concept Model (1.6B parameters), operating on continuous sentence embedding spaces via a Two-Tower diffusion architecture trained on 38.8 billion sentences across 46 languages, dramatically outperforming comparable token-based autoregressive models like Qwen3 1.7B on cross-lingual summarization tasks.
- Parallel Tokenizers: Rethinking Vocabulary Design in Cross-Lingual Transfer of Low-Resource Languages
-
Addressing semantic index fragmentation and severe fertility imbalances in conventional multilingual tokenizers, this paper introduces Parallel Tokenizers, which build monolingual tokenizers aligned at the word level via machine translation and broadcast language identity embeddings to achieve superior cross-lingual transfer across 13 low-resource languages.
- Plasticity without Collapse: Plasticity-Inducing Control Optimizer for Cross-Lingual Continual Pre-Training
-
PICO rethinks continual pre-training as an optimizer-level control problem, combining coordinate-wise utility gating to protect key parameters, periodic spectral sensitivity monitoring to detect implicit degradation, and composed-gate stochastic perturbations to maintain adaptation plasticity, achieving positive forward transfer and the lowest worst-case collapse ratio across multilingual streams without replay buffers or model modifications.
- Polyglot Teachers: Evaluating Language Models for Multilingual Synthetic Data Generation
-
Addressing the common ad-hoc practice of selecting the largest available LLMs as synthetic data generators for multilingual supervised finetuning (SFT), this paper proposes POLYGLOT SCOREโa holistic metric combining intrinsic data quality with extrinsic student downstream performanceโevaluates 10 state-of-the-art models across 6 typologically diverse languages via 1.4M synthesized examples and 240 student models, and demonstrates that model scale does not predict teacher effectiveness while model family matching and method adaptation yield optimal data recipes.
- Predicting Multilingual Classification and Translation Performance of LLMs with Cross-Lingual Alignment โ Is English Enough?
-
This paper comprehensively examines 27 cross-lingual alignment (CLA) variants for predicting LLM performance across classification and machine translation over 44 languages, proposes a target-language independent PMI translation evaluation metric, and demonstrates via convex combination analysis that LLMs predominantly use English as an internal pivot for translation.
- Scaling Unsupervised Word Alignment to Documents via Structural Constraints
-
Addressing the severe performance degradation caused by noise explosion and spurious matches when scaling unsupervised word alignment to full documents, this paper proposes training-free, segmentation-free structural constraint methods (CTFAlign and MDPAlign) that reduce document-level alignment error rate (AER) from 0.412 to 0.326 across six language pairs and significantly enhance downstream translation coverage evaluation and semantic difference recognition.
- Target-Language Generation in Multilingual Models: Activation Steering and Optimal Control
-
This paper formulates cross-lingual target-language text generation as an optimal control problem over hidden activation trajectories, deriving a closed-form Linear Semantic Control (LISECO) projection law and establishing a unified evaluation metricโadhered coherenceโto achieve reliable, prompt-free language forcing across multilingual language models.
- The Blindness of Document-Level Translation Evaluation
-
By introducing a counterfactual evaluation condition (MIX) that recombines segments from multiple systems to disrupt discourse coherence while keeping document layout intact, this paper demonstrates that standard human evaluation protocols (Error Span Annotation) and 14 automatic metrics produce statistically indistinguishable scores, system rankings, and error annotations between coherent and incoherent documents, proving that current evaluation frameworks suffer from protocol blindness rather than annotator perceptual failure.
- The Multilingual FrameNet Corpus
-
This paper introduces the Multilingual FrameNet Corpus (mFNC), the first large-scale harmonized benchmark covering 10 typologically diverse languages (1.5M tokens, ~70k sentences), demonstrating that training discriminative and generative parsers on multilingual frame data dramatically outperforms English-only baselines in both multilingual and zero-shot cross-lingual settings.
- Unlocking Fine-Grained Translation Quality Estimation in LRMs through Mutually Boosting Implicit and Explicit Reasoning
-
Addressing the failure of large reasoning models (LRMs) in fine-grained translation quality estimation due to massive search spaces and fluency illusions, this paper proposes RIEQE, a two-stage framework that decomposes quality estimation into subtasks, pre-trains implicit layer-wise representations via NonThinking-SFT, and then enhances explicit token-wise reasoning via Thinking-RLVR, achieving bidirectional mutual reinforcement and new SOTA performance.
- UrduMMLU: A Massive Multitask Benchmark for Urdu Language Understanding
-
UrduMMLU is the first large-scale, natively authored MMLU-style benchmark for Urdu, featuring 26,389 four-option multiple-choice questions across 26 subdomains and 5 broad domains derived from Pakistani public examination board PDFs and native MCQ banks; evaluation of 30 leading LLMs reveals that while top systems achieve over 97% on STEM, they suffer a dramatic 25 to 40 percentage-point collapse on Urdu-centered humanities and cultural subjects.
- When Does Auxiliary Classification Help Regression in Practice? A Case Study in Arabic Dialectness Modeling
-
Addressing the severe long-tail distribution heavily skewed toward Modern Standard Arabic in continuous dialectness regression, this paper proposes a multi-task learning framework combining 7-bin equal-width discretization with Focal Loss, mitigating negative transfer and improving blind test RMSE on NADI 2024 to 0.1275 (>9% gain).
- When Similar Means Different: Evaluating LLMs on ArabicโHebrew Cognates
-
This paper introduces SemCog Bench, a curated benchmark of 1,858 ArabicโHebrew word pairs and 2,222 contextual sentence MCQs across true cognates, false friends, and loanwords, revealing that LLMs heavily rely on surface-form similarity rather than genuine cross-lingual semantic understanding.