Jonathan Zheng, Zirui Shao, Alan Ritter +1cs.CL cs.LG
Large language models (LLMs) rely on static pretraining corpora, causing their knowledge to become outdated over time. Existing approaches for evaluating knowledge edits either suffer from rapid contamination or rely on counterfactual edits that conflict with rigid existing knowledge. In this work, we propose a synthetic, simulation-driven framework for studying knowledge insertion in LLMs. We introduce {\sc ParallelEvents}, a benchmark of fictional yet realistic future worlds that generates coherent event trajectories for controlled evaluation, avoiding contamination while preserving consistency. Building on this dataset, we develop {\sc Synapse}, a training framework that uses model-generated data to update model parameters via mid-training and instruction tuning. This synthetic pipeline enables scalable knowledge integration without costly human-curated data. Empirically, {\sc Synapse} outperforms existing methods by 14.23\%, demonstrating that simulation-based synthetic training leads to robust and coherent knowledge insertions.
We study self-modeling: an LLM's ability to answer questions about its own behavior. We focus on verifiable behavioral questions, such as whether a prompt edit would change the model's final answer. To measure this capability, we introduce a benchmark that tests diverse types of self-modeling questions. Current models show non-trivial but limited self-modeling skill, and make systematic mistakes on simple counterfactual questions about their own behavior. To improve self-modeling skill, we develop a scalable synthetic-data pipeline that produces self-modeling training data, and show that reinforcement-learning can improve aggregate self-modeling skill across three open-source model families with some transfer to held-out tasks. These gains, however, do not seem to constitute introspection consistently: improved self-modeling may not arise from privileged access to the model's internal decision process.
Synthetic data generation has become a cornerstone for advancing large language models. However, the lack of the quantitative analysis for error tolerance became a critical bottleneck. Consequently, current filtering strategies fluctuate between two extremes: they are either overly aggressive, risking the exclusion of potentially valuable samples, or overly permissive, failing to eliminate erroneous samples effectively. To bridge this gap, this paper introduces Atomic Tree Operation Modeling (ATOM), a framework that decomposes data into functional units ($f(x)\rightarrow y$). ATOM distinguishes benign Operand $x$ perturbations from fatal Operator $f$ perturbations. The former are needlessly discarded by aggressive filtering, while the latter slip through permissive filtering. Our experiments reveal a double dissociation: models are robust to operand perturbations but collapse under operator perturbations. By prioritizing operator over aggressive operand precision, our ATOM-synthesized data outperforms rigorous baselines (e.g., +3.1% gain over LIMA), suggesting that operator diversity matters more than operand precision. Our code is available at https://github.com/Lut-hub/ATOM.
Financial question answering (QA) has emerged as a key benchmark for evaluating the performance of Large Language Models (LLMs) on domain-specific tasks involving complex data formats such as tables, charts, and rich textual narratives. While recent advancements have enabled models to reason across modalities and perform multi-step arithmetic operations, limitations remain in performance consistency, and evaluation reliability. In particular, standard evaluation metrics like Exact Match (EM) often fail to account for minor variations such as differences in units or formats, misleading performance assessments. In this work, we propose a comprehensive pipeline for improving financial QA systems through high-quality synthetic data generation and fine-tuning of smaller language models (SLMs) using Quantized Low-Rank Adaptation (QLoRA). Our pipeline includes aggressive data validation for synthetic question answer generation to ensure the relevance and correctness of synthetic question-answer pairs. We introduce a novel evaluation metric that matches answers computed from arithmetic expressions rather than ground-truth answers; providing a more accurate reflection of model reasoning capability. Furthermore, we propose a modified loss function that aligns predicted and reference expressions using semantic similarity, our novel evaluation metric and standard cross-entropy, resulting in improved performance. Experimental results on benchmark datasets, ConvFinQA demonstrate significant gains in QA accuracy after fine-tuning using synthetic dataset and proposed loss function.
Qiankai Xu, Qiguang Chen, Zixin Su +4cs.CL cs.AI cs.LG
A recent line of synthetic-data work reconstructs the thinking behind existing text rather than rewriting the text itself, but it operates on short web passages, recovers only local thoughts, and leaves the structure of whole documents untouched. Scientific papers are written to a clear and largely uniform structure and make a natural substrate for lifting this paradigm to the document level. We present a pipeline that unfolds each paper into a multi-turn generation trajectory in which a teacher model reconstructs the writing process of the whole paper: a writing request, a global plan, and pre-writing deliberation for each section. All section texts and the abstract are kept verbatim from the source paper. We apply the pipeline to quality-filtered arXiv papers and obtain a corpus for continued pre-training (CPT) that is roughly twice the size of the source text. The same reverse construction extends to instruction data and evaluation. Treating real paper text as the answer yields an SFT dataset. Anchoring tasks in held-out papers yields PAW-Bench, an academic-writing benchmark whose tasks carry their own rubrics and checklists. In controlled experiments CPT on our corpus followed by supervised fine-tuning on public datasets improves writing benchmarks broadly while preserving general reasoning and improving long-document reading. The writing gain persists even when every model is fine-tuned on a dedicated writing SFT dataset. Mixing our SFT data into that recipe lifts academic writing further.
Retrieval-augmented generation (RAG) improves large language models by incorporating external knowledge without retraining, but existing methods often underuse the relational structure encoded in knowledge graphs. Graph-based RAG can capture entity relationships, yet supervised graph retrieval typically requires labeled question-answer data that may not be available for newly constructed graphs. We address this limitation with SelfGraphRAG, a framework that generates question-answer pairs directly from knowledge graph structure and uses them to train a query-conditioned graph retriever. The generated questions capture multi-hop paths and local neighborhoods, providing relational supervision without manual annotation. Experiments on multi-hop question answering and classification benchmarks show that SelfGraphRAG improves retrieval precision and downstream reasoning performance over embedding-based baselines. These results suggest that knowledge graph structure can provide useful supervision for training graph retrievers when labeled data are unavailable.
This paper proposes methods to extract over 50 types of events from a Dutch historical corpus spanning the 17th and 18th centuries. The methods we propose aim to tackle the impossible: extracting the long-tail of the long-tail. Historic data from before the 19th century is in itself a niche domain not covered in the pre-training of Large Language Models, and we aim to extract events only very scarcely annotated in the training data available for this domain. We propose creating expert classifiers for subgroups of the events present in the training data. We make these groupings based on similar frequency in the training data or on semantic relatedness. Experts trained on underrepresented events are assigned higher priority when predicting to avoid being dominated by frequency biases. We refer to this new way of combining classifiers, specifically tailored to protect the long-tail, as ROBE: Reversed-Order-Biased-Experts. We also propose a controlled method to create domain-specific synthetic data. Our two implementations of ROBE outperform a simple fine-tuned encoder model with a .10 increase in recall and a .16 increase in precision respectively. The best model achieves a .10 increase in f1 for a group of long-tail classes in our niche data set.
We present Aslema, our system for NADI 2026 Shared Task 5, which consists of two subtasks: intent recognition and slot filling. We evaluate four omni LLMs in a zero-shot setting and compare them with fine-tuned models. Our results show that fine-tuning consistently outperforms zero-shot inference. We further explore synthetic data augmentation by using an LLM to generate culturally grounded Tunisian Derja utterances, followed by voice cloning to generate synthetic speech. Incorporating this synthetic data improves performance on both tasks. Our final submitted system, based on Qwen3-Omni-30B and trained with a mixture of original and synthetic data, achieves 86.8% intent accuracy and 34.7 WER on the devtest split. On the official test set it ranks 1st in slot filling (59.5 CoER) and 4th among 8 teams in intent recognition (66.1% accuracy). We release our experimental scripts and will soon share the synthetic dataset to support further research in this area.
The rapid progress in Artificial Intelligence has largely bypassed African languages, creating a digital divide that limits AI adoption on the continent. Recent open-source LLMs systematically underperform on African machine translation, while the lack of large-scale, high-quality, open-source parallel data has constrained the development of competitive small language models (SLMs). We introduce *TranslatePsy-AfriSLM*, a collection of open-source MT resources for 19 Sub-Saharan African languages, including curated parallel data, African-specialized synthetic data, and a family of fine-tuned SLMs. Our empirical study shows that unified quality-estimation filtering removes up to 96% of training tokens without degrading quality, and that filtered synthetic data dominates the quality-efficiency Pareto frontier. Fine-tuned on the resulting data mixture, TranslatePsy-AfriSLM outperforms substantially larger systems, including TranslateGemma-27B and Qwen3.5-122B-A10B, with as few as 0.8B parameters.
Language-specific competency (LSC) is the phenomenon of a language model performing better or worse depending on the language of the prompt. In other words, a language model outputs different (and potentially incorrect) responses to the same semantic query when prompted in different languages. Prior work attributes this to an internal misalignment of semantic representation across languages. Currently, there are two main approaches to address LSC in the literature: (1) routing all queries through English, improving performance, but limiting language expressivity to English; or (2) training on language-balanced data, equalizing model performance across languages, but reducing overall performance. In this work, we take a data centric perspective and introduce HOTFIXR: Hardness Optimized Training data For Improving X-Lingual Reasoning. It is a data generation framework that uses models to probe and learn a student model's multilingual weaknesses, and generates data to mitigate them. HOTFIXR can generate multilingual synthetic training data that can improve multilingual performance. We evaluate on three in-distribution tasks, three out-of-distribution tasks, and four out-of-distribution languages. On average, HOTFIXR (1) improves in-distribution performance by 6.2%, (2) reduces catastrophic forgetting (induced by fine-tuning) on OOD tasks by 3.7%, and (3) on OOD languages by 7.1%. Overall, as many real-world applications requires multilingual LLMs, our work contributes to the efforts of making LLMs multilingually proficient. We will release code upon acceptance.
Large language models (LLMs) can generate synthetic training data for text classification, but the quality of generated samples is heterogeneous: some fall in correct class regions of the embedding space while others land in peripheral or cross-class zones. We propose a geometric filtering framework that evaluates each LLM-generated sample by its Euclidean distance to real class examples in a sentence embedding space, selecting only geometrically consistent candidates. A soft weighting mechanism transforms filter scores into sample weights for classifier training. Evaluated across 13 datasets, 5 classifiers, 10 augmentation methods, and over 6,700 configurations, our method achieves +2.61 percentage points (pp) over SMOTE ($p<0.0001$, Cohen's $d=0.95$, 88.9% win rate). The approach generalizes to named entity recognition (+9.26pp, 100% win rate) without filter modification, and is robust across 5 LLMs from 4 providers. A key finding is that the simplest distance-based filter consistently outperforms complex multi-criteria alternatives.
In open-domain human-computer interaction scenarios, large language models (LLMs) frequently encounter user queries that are ambiguous or incomplete. In such cases, directly producing an answer often leads to overgeneralized, erroneous, or low-information responses. In contrast, asking clarifying questions can substantially improve interaction quality. However, existing approaches still rely heavily on manually annotated data or preference alignment to address two fundamental challenges: when clarification is necessary, and which aspect of the query should be clarified. This reliance incurs high annotation costs and limits generalization. To address these challenges, we propose CLAIM, an uncertainty-driven framework for active clarification learning in open-domain settings. CLAIM eliminates the need for explicit human preference annotations by quantifying query uncertainty through the entropy induced by answer disagreements across multiple models. This uncertainty signal is then used to construct high-quality synthetic data, enabling the training of a unified clarification decision model through a combination of supervised learning and reinforcement learning. Specifically, we propose an entropy-driven synthetic data generation pipeline that integrates entropy-based uncertainty estimation with semantic clustering and reasoning-based judgments, enabling reliable automatic annotation of clarification requirements. To train CLAIM, we formulate the clarification process as a structured decision generation problem and adopt a training paradigm that combines supervised fine-tuning (SFT) with group-relative policy optimization (GRPO). Experimental results demonstrate that CLAIM can learn stable and generalizable clarification strategies without relying on manually labeled data, offering a low-cost and robust solution for proactive understanding in real-world open-domain interactions with LLMs.
Javier Rodriguez-Juan, Hiba Arnaout, Jose Garcia-Rodriguez +2cs.CL cs.LG
Synthetic generation of Cognitive Behavioral Therapy (CBT) sessions is challenged by two competing demands: adhering to strict therapeutic structure while modeling the resistant, unpredictable behavior of real patients. Existing script-based methods fail to capture dynamic therapeutic interactions, while multi-agent approaches struggle to adhere to CBT's sequential structure; both suffer from sycophancy, producing overly compliant patients that misrepresent real clinical settings. In this work we introduce ODRA, a novel framework for synthesizing therapy dialogues through a Chain-of-Thought (CoT) strategy grounded in foundational CBT guidelines (Beck, 2020). ODRA further incorporates a resistance orchestrator to solve patient sycophancy, which employs steering techniques to elicit behaviors aligned with their resistance level. Automated and expert evaluations show that ODRA significantly outperforms existing methods across therapeutic skills, CBT alignment, and patient behavioral fidelity, with licensed psychologists preferring ODRA sessions across 12 of 13 clinical metrics. Furthermore, models fine-tuned on our dataset demonstrate superior therapeutic performance against both cooperative and resistant patients, validating that explicit resistance modeling in synthetic training data directly translates to downstream clinical robustness.
Synthetic textbook data has improved language model pre-training, but prior work largely treats the benefit as a property of generated content or local rewriting style. We study a different factor: whether related content is organized into coherent book-level documents. We contribute both a scalable synthesis pipeline and controlled evidence that this organization matters. The pipeline retrieves source material from a pre-training corpus, clusters it into topical units, plans hierarchical tables of contents, and assembles source-grounded sections into complete books (our Full setting), yielding 686K textbooks (32B tokens) across 15,000+ disciplines. Replacing natural books in a mid-training mix with this corpus improves downstream performance by +1.09 on average. Controlled comparisons then disentangle the relevant design factors. A content-matched Split condition holds generated text and tokens fixed but treats each section as an independent document; Full's +1.02 mean gain isolates document packaging. A length-matched RandomConcat control that joins sections from different books remains below Full, ruling out document length alone. A retrieval-pool-matched Rephrase condition independently rewrites individual retrieved documents under the same audience-by-style scheme, without clustering, TOC planning, or book assembly; Full's +1.17 gain demonstrates the value of structured synthesis. On Llama3-8B, Full likewise outperforms both RandomConcat and Natural Books, supporting book-level organization as a useful axis for synthetic pre-training data design.
Varun Ghat Ravikumar, Sina Ahmadi, Lena Jäger +1cs.CL
Most endangered languages lack the parallel data required for machine translation, despite the existence of descriptive grammar books. We introduce a pipeline that uses large language models to extract grammatical rules, example sentences, and lexicons from grammar books and generate synthetic parallel corpora for fine-tuning-rather than feeding grammar content into prompts at inference time, as in prior work. Validated on three typologically diverse low-resource languages-Kalamang (Papuan), Tuatschin (Romance), and Mandan (Siouan)-we show that fine-tuning on synthetic data improves over seed-data baselines in 75% of configurations for Kalamang and 59% for Tuatschin, with best-case ChrF++ gains of +8.8, +5.3, and +3.3 respectively. Through a systematic factorial study across 96 configurations varying target part-of-speech, retrieval granularity, and sample volume, we identify which factor combinations drive gains and where they break down. Our results demonstrate that static linguistic documentation can be repurposed for machine translation fine-tuning, offering a practical path towards translation tools for severely under-resourced languages.
Model collapse is a central challenge in learning from synthetic data: as later-generation large language models (LLMs) are trained on an increasing proportion of model-generated data, performance can degrade due to narrowed coverage and accumulated bias. Existing work mainly studies how to bound this degradation. In iterative model evolution, however, the more meaningful objective is to ensure that each successive model improves over its predecessor, which requires diagnosing collapse at a granularity that is actionable for data curation. We study this problem in synthetic data self-improving for instruction tuning. We show that collapse in this setting is not simply uniform performance degradation, but can appear as a polarization of competence, where synthetic training reinforces already strong skills while further degrading weak ones. Motivated by this observation, we propose KITE (Knowledge-boundary Instruction Tuning via Exploration), a two-stage framework that combines failure-guided data generation with boundary-aware uncertainty curation. Experiments across several datasets and multiple open-source LLMs show that KITE yields more stable improvement than strong synthetic-data baselines.
Rubrics provide structured, fine-grained signals for training and evaluating large language models (LLMs). Yet reliable query-specific rubrics are difficult to construct. Existing approaches often derive supervision from human-written rubrics, preference data, or sampled responses. Direct query-to-rubric generation avoids these resources, but provides no explicit check that a plausible rubric is useful. Such a rubric may fail to distinguish answer quality, reward an optional style, or penalize a valid alternative strategy. We introduce Rubrics on Trial, a query-only framework that evolves a rubric set from an empty set without external annotations or model training. It derives supervision solely from synthetic rubric-conditioned response pairs and validates each proposed rubric before adding it, screening out non-discriminative, over-specific, and style-only candidate rubrics. Experiments across five preference benchmark suites demonstrate the effectiveness of Rubrics on Trial, which achieves the best average accuracy and leads on six of seven evaluation sets.
Text-to-SQL is a fundamental task in natural language processing that enables users to interact with structured databases using natural language. While large language models (LLMs) have demonstrated remarkable performance on this task, their substantial computational requirements hinder deployment in resource-constrained settings. In this paper, we introduce SQuaD-SQL (Small-Qualified and Distilled for SQL), a novel approach that empowers small language models (SLMs) to approach the performance of LLMs on the Text-to-SQL task while significantly improving efficiency through knowledge distillation and synthetic data generation. Our method comprises three key components: (1) LLM-based synthetic data generation, where structured knowledge is extracted from LLMs via carefully designed prompting strategies; (2) parameter-efficient fine-tuning, enabling full model training on a single consumer-grade GPU; and (3) domain-adaptive fine-tuning, where domain-specific synthetic data further enhances performance in targeted domains. Experiments on the WikiSQL dataset demonstrate that SQuaD-SQL achieves an execution accuracy of 86.9% on the test set, approaching the performance of LLMs while offering faster inference and lower memory usage. These results suggest that, with proper training strategies, SLMs can serve as practical and efficient alternatives for Text-to-SQL applications in resource-limited environments.
Andrea Scarinci, Virginia Negri, Brayan Impata +3cs.CL cs.AI
Fine-tuning large language models (LLMs) for e-commerce attribute extraction requires labeled data representative across thousands of product types, attributes, and multiple languages. This combinatorial scale translates to millions of annotations, rendering human labeling prohibitively costly. While recent work has demonstrated synthetic label generation using LLMs, deploying such approaches at industrial scale requires integrated quality control mechanisms. We present SynthAVE, a large-scale human-validated benchmark for attribute value extraction spanning 12,726 products across 229 product types, 792 attributes, and 4 languages (Spanish, French, Italian, German). To validate synthetic labels at scale, we introduce a multi-LLM arena framework where samples are independently evaluated by 21 judge configurations (7 model families $\times$ 3 prompts), with final labels determined via majority voting. The majority vote ensemble agrees with human experts at Cohen's $κ= 0.92$ (95.2% agreement), while individual judges show substantial inter-model agreement (Fleiss' $κ= 0.76$). This demonstrates that diverse models with varying individual judgments aggregate into highly reliable predictions, enabling cost-effective validation at scale while maintaining quality parity with human review.
Modern data-driven marketing relies on large amounts of consumer data, yet collecting such data can be costly, time-consuming, and difficult to scale. This research examines whether large language models (LLMs) can be used to generate synthetic consumer data for projective techniques, a set of methods designed to elicit consumer associations, emotions, wants, and needs. We test LLM-generated responses across multiple projective tasks, LLMs, prompting strategies, and temperature settings, and compare them with human responses from a primary research study on perceptions of city tourism destinations. Human and LLM responses were analyzed using linguistic measures, diversity and concentration metrics, topic models, and top-term analyses. The results show substantial overlap between human and LLM responses in broad topics and associations, but also important differences in style, linguistic structure, and the way diversity is generated. Recommendations are given on how to best utilize LLMs for generating synthetic consumer data, how model and prompt choices shape response quality, and on recognizing the limitations of LLM synthetic consumer data generation.
Synthetic data can be scaled along two routes: Source Expansion (SE), which enlarges the source by adding seed materials or generators, and Fixed-Source Synthesis (FSS), which holds the source fixed and scales the generation budget. Existing scaling studies typically expand the source as the data grows, conflating SE with FSS and leaving FSS underexplored. We isolate FSS by holding the seed-question pool and teacher model fixed, varying only the per-question response budget under Rejection Sampling (RS). We adapt the rectified scaling law to FSS, deriving it from how repeated sampling covers a fixed source. Empirically, the derived form, fit on low budgets, predicts performance at the held-out highest budget for every evaluated teacher--student pair. At matched total-sample budgets, SE and FSS are comparable at small budgets; at large budgets, adding seed questions outperforms spending the same budget on more responses. Within FSS, however, neither synthesizing additional questions from the existing seeds nor varying the synthesis protocol outperforms plain RS at matched budgets. FSS is thus a bounded scaling axis and a controlled setting for comparing synthesis protocols. We will release our code and data to facilitate further research.
Maximilian Idahl, Jörg Tiedemann, Sampo Pyysalo +19cs.CL
Open web-scale pre-training corpora remain concentrated in English, limiting multilingual LLM development. We introduce MultiSynt/MT, an open synthetic parallel corpus with approximately 4.8 trillion target-language tokens across 36 European languages, produced by translating 100 billion high-quality Nemotron-CC tokens with Tower+ and OPUS-MT/HPLT-MT systems. For many medium- and lower-resource European languages, this is the largest openly available pre-training resource. On a broad multilingual benchmark suite, reference LLMs trained on MultiSynt/MT reach the final score of HPLT 2.0, a native-data baseline, using roughly 72% fewer pre-training tokens, and outperform it by approximately 15% relative at a matched 100B-token training budget. Our analyses also identify evaluation blind spots: standard multiple-choice benchmarks miss translation-quality differences that a fluency-sensitive LLM-as-judge evaluation cleanly recovers on the trained LLMs (with no fluency deficit in MultiSynt itself), and Norwegian idiomatic and culturally grounded tasks remain better served by native data. We release the corpus, including row-aligned translations from multiple systems, to support controlled research on multilingual pre-training data and evaluation.
Maltese has substantial text corpora and pretrained language models, but paragraph-scale OCR training data remains scarce; NOMOCRAT provides 57 verified annotated pages. LV-ROVER-MLT combines synthetic fine-tuning of Tesseract~5 with five complementary recognition streams and lexicon-gated word-level arbitration adapted to Maltese diacritics and hyphenation. In the DocEng~2026 Maltese OCR competition, the system placed first with held-out CER 0.0074; the next-ranked submission scored 0.0161 and NOMOCRAT scored 0.0163. The same approach produced a significant improvement over stock Tesseract on Luxembourgish, while the Hungarian result was inconclusive. A 36,803-pair Maltese OCR corpus constructed from EUR-Lex and Wikipedia provides an additional paragraph-level resource. Code, model weights, and corpus data are public.
Language models are increasingly taught from synthetic question--answer (QA) supervision: a model generates questions about a document, answers them from the same text, and the resulting pairs are used to fine-tune, distill, or compress knowledge into another model. We show that this generation step is not neutral preprocessing. It is an implicit policy that both selects which evidence becomes training signal and decides how that evidence is answered, and it is fragile at both stages. When choosing what to ask, generators do not scan a document uniformly. Coverage saturates early and concentrates on salient spans, diverse prompts converge on the same regions, and what looks question-worthy is driven by local presentation. As a result, salient artifacts such as poorly cleaned markup can hijack question generation across model families and scales. When answering, the model that produces the supervision tends to obey instruction-like passages embedded in the text. This compliance depends on the intent and surface form of the passage rather than its strictness, and is worst under task conflict, where larger models comply more often. These failure modes arise from choices made during QA generation, so they can be reduced without changing the training loop. Tying each question to a fixed target reduces biased selection, and filtering instruction-like spans before answering lowers mean injection compliance from $88\%$ to $13\%$ in our evaluation while retaining nearly all clean text.
Roland Roller, Vera Czehmann, Derya Erman +13cs.CL
Conversational data collected in domains such as healthcare or social sciences is a valuable resource for research and automated analysis. However, responsible data sharing requires the detection and removal of personally identifiable and sensitive information to protect individual privacy. To support the development and evaluation of automatic de-identification systems, we present DialogPII, a multilingual dataset of synthetic dialogs and speech-derived transcripts for personal information detection. DialogPII covers eight interaction scenarios (emergency calls, medical anamnesis interviews, therapy sessions, insurance communication, customer support, clinical interviews regarding an AI-supported dashboard, police reports, and group therapy discussions), 19 entity types, and 11 languages (English, Arabic, Finnish, French, German, Hindi, Italian, Polish, Portuguese, Spanish, and Turkish). Dialogs were generated semi-automatically using large language models, manually curated for plausibility and diversity, and localized to country- and city-specific contexts. All dialogs were additionally converted to speech via text-to-speech synthesis, transcribed with Whisper, and annotated through automatic projection and manual correction, yielding aligned written and speech-derived resources across all languages. We further release baseline multilingual named entity recognition models and provide technical validation through inter-annotator agreement analysis, translation quality evaluation, annotation projection assessment, and benchmark experiments with transformer-based sequence labeling models.
Knowledge injection via synthetic data is crucial for enhancing Large Language Models (LLMs). However, current synthesis methods simply stop at preset token counts or fixed data ratios, lacking awareness of knowledge distribution. This results in some domains being sparse while others are redundant, limiting LLM knowledge boundaries. We revisit knowledge injection from a distribution perspective and hypothesize that an optimal knowledge distribution exists to maximize knowledge boundary expansion. We propose KDoS (Knowledge Distribution-optimized Synthesis), a framework that introduces knowledge density to drive synthesis through a three-stage feedback mechanism, shifting from blind generation to distribution-optimized synthesis. We construct Wikipedia-based synthetic data with varying knowledge distributions and conduct experiments on models from 0.6B to 16B (Qwen, Ling, LLaMA) and data scales from 1B to 5B tokens. Our key findings are: (1) an optimal knowledge distribution consistently maximizes boundary expansion; (2) this distribution is stable across backbones and scales; (3) KDoS outperforms baselines across six knowledge benchmarks. Our work offers a new perspective and practical framework for synthetic data-driven knowledge injection.
Data curation is a critical part of post-training pipelines for large language models, yet existing tools often treat ingestion, deduplication, synthetic generation, and quality filtering as separate stages. This fragmentation makes it difficult to audit pipeline decisions or understand why individual samples are rejected. CuratorKIT is an open-source Python library that covers this full lifecycle in a single configurable pipeline. The framework is composed of six source format readers and automatic schema detection, a pre-generation data hygiene layer for credentials, PII, and toxic content, eight LLM-powered generation tasks, three complementary quality gates with provenance-exact hallucination verification, structured adaptive recovery, and five training-ready export formats compatible with TRL, Unsloth, and AlignTune. Every pipeline decision is recorded in an append-only per-sample provenance chain, and rejected samples carry structured failure reasons rather than being silently discarded. CuratorKIT supports 100+ LLM providers through LiteLLM, exposes both a Python API and a YAML-driven CLI, and is designed for practitioners who need reproducible, auditable data pipelines at scale .
Generating high-utility synthetic data for intent classification typically requires human-annotated seed data, which is often unavailable in fast-paced industrial settings. In this paper, we propose a framework for synthetic dialogue generation that works entirely without human-annotated data, relying solely on intent definitions. Our proposed dialogue generation framework utilizes two different types of topic and style attributes to improve data diversity. Also, we propose two novel post-hoc stylization models called Univ and Exam to transform synthetic LLM-generated utterances into more varied, human-like linguistic styles. To enhance data quality, we utilize an LLM-as-a-judge filtering process. Experimental results on both industrial and public datasets demonstrate that the proposed approach achieves up to 93.3% of the performance obtained using human-annotated training data. Crucially, the findings reveal that style diversity is more critical than topic diversity for synthetic data utility, as it prevents models from learning spurious stylistic correlations. Furthermore, the study shows that incorporating style attributes during the generation process is more effective than post-hoc style adaptation.
Driven by massive amounts of web-scale data, generative AI (GenAI) has achieved remarkable progress, enabling various applications in diverse sectors. The advances of GenAI have actuated practitioners to use AI-synthesized data for training next-generation AI models. Undeniably, using synthetic data has alleviated the increasing stringent demand for data supply. Unfortunately, it also introduces a new critical issue: in a self-consuming cycle between model and data, the model ultimately collapse, raising more trustworthiness concerns to GenAI. In recent years, increasingly more studies have investigated the phenomenon of model collapse (MC) and explored potential solutions to mitigate it. However, the review of the phenomenon of MC still remains blank. To fill this gap, this paper provides an up-to-date overview of these studies for consolidating and reviewing the progress of MC in different application scenarios and countermeasures for mitigating MC. We also highlight challenges and future research opportunities.
Large instruction-following models are powerful but costly to deploy, particularly in finance, where labelled data are limited by confidentiality and expert annotation cost. We present an efficient framework for financial sentiment analysis through distillation with synthetic data, transferring knowledge from a large instruction-tuned teacher to compact student models. The framework is designed for low-resource conditions, where a small set of real examples are collected and labelled by hand. The framework then clusters the examples and uses the clusters to select seeds for generating synthetic examples via structured few-shot prompting. Experiments show that clustering-based seed selection yields more representative synthetic data than random sampling, enabling compact models to achieve strong performance with minimal supervision. Notably, on a more complex and noisy text domain, the compact model trained on the complete synthetic-seed corpus even outperforms the teacher model, while remaining competitive on formal text. The framework provides a practical route toward resource-efficient domain adaptation in financial NLP with minimal human labelling effort.