The results show that curriculum-guided structured supervision can transfer multi-component tutoring behavior to a sub-1B model under these resource constraints.
Abstract
Bangla-English tutoring requires more than producing a correct translation: learners also need explanations of grammar differences, awareness of their likely errors, and targeted practice. We present TRACE-BN, a curriculum-guided dataset of structured tutoring traces for Bangla-speaking learners of English at the CEFR A1-A2 level. Each trace combines word-level glosses, literal and natural translations, Bangla grammar explanations, a plausible learner error, and a targeted practice question with its answer. The traces are generated by Gemini 3.5 Flash Lite as the teacher model from NCTB Classes 9-10 English curriculum units, then filtered for structural validity, script integrity, and semantic duplication. We transfer the resulting structured tutoring behavior to Qwen3-0.6B using LoRA with 4-bit quantization for resource-constrained offline deployment. On held-out inputs, schema validity increases from 85.4% to 95.8%, while, against teacher-model references, chrF++ improves from 15.28 to 34.77 and BLEU from 4.52 to 21.03. Field-level evaluation by two independent judges shows improvements across translation, grammar explanation, learner-error diagnosis, and practice alignment, while a human audit supports the quality of the supervision data. The results show that curriculum-guided structured supervision can transfer multi-component tutoring behavior to a sub-1B model under these resource constraints. The dataset, model checkpoints, and code are publicly available at https://huggingface.co/datasets/RaiyanKhaan/Trace-BN
Mathematical reasoning remains challenging in low-resource languages such as Bangla. We study whether teacher-generated Bangla Chain-of-Thought (CoT) supervision provides benefits beyond ordinary supervised fine-tuning. We construct \textsc{MathShikkha}, a Bangla mathematical reasoning dataset with GPT-5.4-generated rationales, and fine-tune four 4B--7B student models under a matched protocol in which answer-only and CoT conditions share data splits, response-only loss masking, decoding, and scoring, differing only in the training target. In-domain, CoT provides no significant improvement over answer-only fine-tuning for three stronger backbones (paired bootstrap 95\% CIs include zero; exact McNemar $p \geq 0.17$), despite generating 15--52$\times$ more tokens, but significantly improves the weaker 4B model by 18.56 points ($p<0.0001$). On the larger, contamination-audited BanglaMATH benchmark, this pattern reverses: CoT significantly outperforms answer-only supervision for all four models by 20.1--28.1 points (all $p<0.0001$). Answer-only fine-tuning also reduces out-of-domain accuracy below the base model for three models, whereas CoT preserves or improves it for all four. A human study with two co-author annotators, external-expert adjudication, and Cohen's $\kappa = 0.76$--$1.00$ finds no significant CoT improvement over the base model on reasoning-content criteria; instead, its measurable effect is target-language adherence and producing inspectable reasoning. Overall, rationale supervision's value depends on backbone capability and distribution shift: in this setting, its main benefits are Bangla adherence, auditable reasoning, and out-of-domain robustness rather than improved in-domain reasoning validity.
Instruction tuning has become the standard method for adapting large language models to follow human intent, yet existing instruction datasets are dominated by English-language general-knowledge tasks and lack coverage of specialized pedagogical domains. This paper presents IKS-Instruct, a dataset of 24,795 instruction-response pairs for teaching language models to deliver educational content grounded in Indian Knowledge Systems (IKS). The dataset spans seven languages (English, Hindi, Sanskrit, Tamil, Telugu, Kannada, and Malayalam), covers 41 pedagogical techniques from the Vedic oral and mathematical traditions, and is aligned with the Central Board of Secondary Education (CBSE) curriculum for classes 6 through 12. The pairs are derived from six source types: classical text corpora (Bhagavad Gita, Thirukkural, Sangam literature, Vedic texts), curriculum-aligned pedagogical templates, Vedic mathematical sutra demonstrations, bilingual instruction pairs, technique-grounded multi-turn dialogues, and cross-tradition comparative analyses. Quality is assessed through a multi-judge evaluation framework in which independent language models score responses on 12 dimensions including technique fidelity, pedagogical quality, factual accuracy, and IKS cultural depth. Under a uniform five-judge external panel (median aggregation over 1,201 stratified items), the strongest IKS-Instruct fine-tune of a compact 7B model reaches a median judge score of 6.39, within 0.15 of a strong general-purpose reference model (Nemotron-Nano at 6.54) at a fraction of its deployment cost, while the base model without IKS fine-tuning scores near zero on the IKS-specific dimensions. Model quality does not increase monotonically with data curation, a result we report together with the corresponding data-quality gains.
Shwetha Singaravelu, G. Muruganantham, L. Rajendran et al.· arXiv.org· 0 citations
In non-native English classrooms, multimedia resources often produce a cognitive gap of high input but low output: visual contexts and target-language forms fail to form stable semantic connections. To address this problem, this study proposes a three-stage multimedia teaching strategy of context anchoring, modal verification, and output transfer. First, a micro-context video library is constructed using word-vector semantic association, with target vocabulary anchoring the semantic field while 12-second clips control cognitive load. Second, a semi-structured virtual dialogue agent dynamically triggers recast and clarification feedback according to learners’ spoken output, grammar breaks, and pragmatic deviations. Third, digital narrative reconstruction tasks require learners to convert video-based semantic information into written storyboards, promoting cross-modal syntactic transfer. Experiments show that the semantic priming effect size increases to 145.2 ms, speech-flow break density decreases to 12.3, and average clause nesting depth rises to 1.89. The framework turns multimedia into a cognitive scaffold and is compatible with networked multimedia transmission, wireless interactive classrooms, and electromagnetic-safe digital learning spaces.
Juan Wu, Yingying Du, Xiaohui Zhang et al.· Advanced Electromagnetics· 0 citations
Results suggest that code-mixing-aware RAG chatbots can meaningfully improve DSA learning outcomes while remaining feasible for instructor-led deployment in low-resource settings.
Saroj Bhandari, Puja Dhital· Aadim Journal of Multidiscip...· 0 citations
An in-depth evaluation of instruction tuning for Arabic NLP tasks using three prominent LLMs: LLaMA 3.1-8B, AceGPT-v2-8B, and Qwen3-8B shows that instruction tuning consistently improves performance across most tasks, with notable variations in effectiveness across different tasks and prompts.
Maged Saeed Al-shaibani, Zaid Alyafeai, Irfan Ahmad· Language Resources and Evalu...· 0 citations
Evaluation of six frontier models reveals that even the strongest model reaches only 63.1\% pass rate, with many tasks unsolved by any model, highlighting that multilingual coding competence is a distinct and underexplored capability axis.
Yunsu Kim, Kaden Uhlig, Ashwin Purohit et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
