{"id":"ca0c8012-4897-4ba2-82f0-1fb24c9e6e19","arxiv_id":"2606.26015","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Tatoxa outperforms open and commercial LLMs for Tatar text detoxification using a new dataset, with native Tatar training beating cross-lingual transfer from Russian.","lead":"The paper presents Tatoxa, a system for text detoxification in the low-resource Tatar language, plus a new dataset for it. A smart generalist might read it to see how AI safety tools can be built when little native data exists.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"No significant objection identified","rationale":"Reader correctly flags the dataset as unverifiable from the abstract alone; the same limitation applies here, so no new or stronger concern can be formulated. The central claims cannot be stress-tested without the missing experimental details.","tokens_in":1631,"tokens_out":229,"duration_ms":16758,"concrete_test":"Retrieve the full PDF from arXiv:2606.26015 and extract the dataset section plus experimental tables; verify whether the Tatar corpus size, annotation agreement, and prompting/evaluation protocol for commercial LLMs are reported in sufficient detail to support the outperformance claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Full manuscript text is not provided in the query (only abstract and reader's note on inaccessibility), so no concrete technical detail of the argument, dataset construction, evaluation protocol, or LLM comparison setup can be examined. Without access to sections describing the Tatoxa model, the new Tatar dataset (size, annotation process, splits), or the exact metrics and baselines, no load-bearing assumption can be isolated or tested.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper presents Tatoxa, a novel system for text detoxification in the low-resource Tatar language. It introduces a new dataset designed for fine-tuning and evaluation in low-resource settings. Comparative experiments claim that Tatoxa outperforms existing open-source and proprietary commercial LLMs on key quality metrics. Cross-lingual transfer experiments indicate that training on native Tatar data significantly outperforms transfer from other languages, including culturally close Russian, even when using a large Russian corpus.","tokens_in":1680,"tokens_out":417,"duration_ms":25221,"significance":"If the empirical claims are substantiated with rigorous evaluation, this would represent a meaningful contribution to NLP for low-resource languages by providing both a dedicated detoxification system and evidence favoring language-specific data over cross-lingual transfer. The focus on Tatar addresses an under-served language and the safety application is practically relevant. However, the absence of methods, metrics, dataset details, or results tables in the available material prevents assessment of whether the claims are supported.","major_comments":[{"comment":"The abstract asserts outperformance on 'key quality metrics' and superiority of native data over cross-lingual transfer, but no methods section, evaluation protocol, dataset statistics (size, annotation process, splits), metrics definitions, or results tables are provided. This prevents verification of the central empirical claims and the weakest assumption that the new dataset is representative and of adequate quality.","section":"Abstract / Experiments (inferred)"},{"comment":"Without access to the dataset construction details or baseline descriptions, it is impossible to evaluate whether the comparisons to open-source and commercial LLMs are fair or whether the cross-lingual experiments control for corpus size, domain, and model scale appropriately.","section":"Dataset and Experiments sections (inferred)"}],"minor_comments":[],"recommendation":"uncertain","confidential_remarks":"The provided material consists only of the abstract and a note that full text is in an inaccessible cache; a full review requires the complete manuscript with methods, data, and results."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for their detailed review and for highlighting areas where the presentation of our work could be improved. We provide point-by-point responses to the major comments below.","responses":[{"response":"The full manuscript contains a dedicated Dataset section that details the construction process, including size, annotation by native Tatar speakers, and data splits. The Experiments section describes the evaluation protocol, defines the metrics used for quality assessment, and presents results in tables comparing Tatoxa to other models. We will revise the manuscript to include explicit pointers from the abstract and introduction to these sections to facilitate verification.","revision_made":"partial","referee_comment":"The abstract asserts outperformance on 'key quality metrics' and superiority of native data over cross-lingual transfer, but no methods section, evaluation protocol, dataset statistics (size, annotation process, splits), metrics definitions, or results tables are provided. This prevents verification of the central empirical claims and the weakest assumption that the new dataset is representative and of adequate quality."},{"response":"The manuscript provides descriptions of the dataset construction and the specific baselines employed, including both open-source and commercial LLMs. The cross-lingual transfer section explains the experimental controls, including the use of a larger Russian corpus for comparison. We will expand these descriptions with additional details on corpus characteristics in the revised version to allow for a more thorough assessment of fairness.","revision_made":"yes","referee_comment":"Without access to the dataset construction details or baseline descriptions, it is impossible to evaluate whether the comparisons to open-source and commercial LLMs are fair or whether the cross-lingual experiments control for corpus size, domain, and model scale appropriately."}],"tokens_in":1277,"tokens_out":371,"duration_ms":21816,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main point is that this paper claims to deliver the first detoxification system for Tatar plus a new dataset, and reports that training on native Tatar data beats cross-lingual transfer from Russian even when plenty of Russian data is on hand. It also says the Tatoxa approach beats both open-source and commercial LLMs on the metrics used.\n\nWhat it does is extend existing detoxification work to a new low-resource language and provide some evidence that transfer is not always the easy win. That finding is worth having on record for anyone who works on content safety outside high-resource languages.\n\nThe soft spot is obvious from the abstract alone: no dataset size, no annotation details, no exact metrics, no description of the model or baselines. Without those pieces it is impossible to judge whether the comparisons are fair or whether the native-data advantage is real. The assumption that the new dataset is representative enough for the claims is left untested in what is visible.\n\nThis is for people who need coverage in Turkic or other low-resource settings and who might use the dataset if it is released. A reader focused on multilingual safety tools could get practical value from the native-vs-transfer result if the experiments check out.\n\nI would send it to peer review so the methods and results can be examined properly; the language gap it targets is real even if the current write-up is thin on evidence.","headline":"Tatoxa introduces the first reported detox system and dataset for Tatar with native data beating Russian transfer, but the abstract gives no methods or numbers so the claims can't be checked.","tokens_in":2162,"tokens_out":360,"would_cite":false,"duration_ms":34915,"reading_group":"maybe","serious_thinker":"unclear","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"A fine-tuned model trained on native Tatar data outperforms both open-source and commercial LLMs at detoxifying abusive text.","keywords":["text detoxification","low-resource languages","Tatar","cross-lingual transfer","large language models","abusive content","fine-tuning"],"falsifier":"A replication that trains the same architecture on the Tatar dataset and finds equal or lower detoxification quality than a Russian-transferred model on the same Tatar test split would falsify the central performance claim.","tokens_in":2555,"feed_emoji":"🛡️","tokens_out":580,"duration_ms":21205,"temperature":0.7,"pith_summary":"The paper presents Tatoxa, a system for automatically detecting and rewriting harmful content in the Tatar language. It introduces a dedicated dataset for fine-tuning and evaluation in this low-resource setting. Experiments demonstrate that models trained directly on Tatar data achieve higher quality scores than large language models applied zero-shot or few-shot. Cross-lingual transfer from Russian or other languages yields weaker results even when large Russian corpora are used for pre-training.","feed_headline":"Native Tatar data beats Russian transfer for text detox","feed_subtitle":"A new system and dataset show training on Tatar itself outperforms both commercial LLMs and transfer from even closely related languages.","key_machinery":"The Tatoxa detoxification model, obtained by fine-tuning on the newly introduced Tatar dataset of toxic and detoxified sentence pairs.","core_discovery":"Tatoxa is a detoxification system fine-tuned on a new Tatar dataset that detects abusive language and produces safer rewrites. On held-out Tatar test sets the approach records higher scores than both open-source models and proprietary commercial LLMs. Separate transfer experiments show that models initialized from Russian or multilingual checkpoints perform worse than the native-Tatar model despite the availability of substantially larger Russian training data.","pith_inferences":["Similar native-data collection efforts may be needed for other Turkic or agglutinative low-resource languages.","The performance gap suggests that cultural and lexical specificity of abusive language limits cross-lingual generalization.","Future work could test whether the same native-data advantage appears in related tasks such as hate-speech detection."],"forward_implications":["Native-language fine-tuning is required for competitive detoxification performance in low-resource languages.","Large corpora from a related language such as Russian do not substitute for native Tatar data in this task.","The released Tatar dataset supports both model development and standardized evaluation.","Commercial LLMs do not automatically solve detoxification for languages outside their primary training distribution."],"fun_headline_variants":["Tatoxa outperforms LLMs on Tatar detox","Native Tatar beats Russian transfer for detox","New Tatar dataset outperforms transfer learning","Tatoxa detox system beats commercial LLMs"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The newly created Tatar dataset accurately represents real abusive language and produces reliable quality measurements for comparison against LLMs.","fun_headline_variants_meta":{"raw":{"variants":["Tatoxa outperforms LLMs on Tatar detox","Native Tatar beats Russian transfer for detox","New Tatar dataset outperforms transfer learning","Tatoxa detox system beats commercial LLMs"]},"model":"grok-4.3","cost_usd":0.005429,"raw_usage":{"total_tokens":2567,"prompt_tokens":576,"num_sources_used":0,"completion_tokens":52,"cost_in_usd_ticks":54287000,"prompt_tokens_details":{"text_tokens":576,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1939,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":576,"tokens_out":52,"duration_ms":13172,"temperature":1.0,"reasoning_tokens":1939,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-25T19:15:50.338364+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A replication that trains the same architecture on the Tatar dataset and finds equal or lower detoxification quality than a Russian-transferred model on the same Tatar test split would falsify the central performance claim.","supporting_citations":[],"review_version":1}