{"paper_id":"2502.16023","title":"Contrastive Similarity Learning for Market Forecasting: The ContraSim Framework","authors":["Nicholas Vinden","Raeid Saqur","Zining Zhu","Frank Rudzicz"],"paper_date":"2025-02-22","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2502.16023v1","page_url":"https://thequant.space/flowcharts/contrastive-similarity-learning-for-market-forecasting-the/","scores":{"empirical_rigor":7,"math_complexity":3.5,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper proposes ContraSim, a Contrastive Similarity Space Embedding Algorithm that generates headline augmentations with similarity scores and uses Weighted Self‑Supervised Contrastive Learning to create a weighted embedding space.","quote":"We introduce the Contrastive Similarity Space Embedding Algorithm (ContraSim), a method that generates headline augmentations with meaningful and nuanced similarity coefficients.","summary":"We introduce the Contrastive Similarity Space Embedding Algorithm (ContraSim), a novel framework for uncovering the global semantic relationships between daily financial headlines and market movements. ContraSim operates in two key stages: (I) Weighted Headline Augmentation, which generates…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Headlines from WSJ and other sources are representative of daily market movements.","type":"data","stated":"implied","quote":"... collected and concatenated together alongside the movement of the US equities market (ticker: $SPY) for the corresponding day."},{"assumption":"Semantic similarity between headlines correlates with similar market direction.","type":"model","stated":"implied","quote":"WSSCL inherently groups headlines associated with similar market directions closer"}]},"datasets":{"status":"available","items":[{"name":"NIFTY‑SFT","provider":"WSJ","kind":"text-news","access":"public","period":"2010/01/06-2017/06/27","universe":null,"quote":"... alongside the movement of the US equities market (ticker: $SPY) for the corresponding day."},{"name":"BigData22","provider":null,"kind":"text-news","access":"public","period":"2019/07/05, 2020/06/30","universe":null,"quote":"likewise is a financial news headline dataset, but news headlines are composed of tweets as apposed to WSJ headlines."},{"name":"IMDB Review","provider":"IMDB","kind":"text-news","access":"public","period":"1874, 2020","universe":null,"quote":"... a collection of human‑written reviews for a list of movies alongside the movie’s overall review score."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"On the NIFTY‑SFT dataset, the combined ContraSim + LLM model achieves a balanced accuracy of 37.79%, a 13% increase over the baseline and a 7% improvement over LLM‑only embeddings.","kind":"performance","quote":"applying this conjunctive approach to the NIFTY‑SFT dataset results in a balanced accuracy of 37.79%, reflecting a 13% increase over the baseline and a 7% improvement compared to using only LLM embeddings."},{"finding":"On the IMDB dataset, the composite model improves accuracy by 6.8% and F1 score by 0.0496 relative to the baseline LLM.","kind":"performance","quote":"the composite model outperformed the baseline LLM, achieving a 6.8% increase in accuracy and a 0.0496 improvement in F1 score."},{"finding":"ContraSim’s embedding space consistently outperforms the baseline on density metrics (g‑KNN, KNN, KL‑Divergence, JSD) across all three datasets.","kind":"statistical","quote":"ContraSim embedding space, optimized through WSSCL and CWCL losses, consistently outperforms the baseline in g‑KNN, KNN, KL‑Divergence, and JSD metrics."},{"finding":"Weighted Self‑Supervised Contrastive Learning naturally clusters headlines that correspond to similar market directions, even without explicit labels.","kind":"theoretical","quote":"WSSCL inherently groups headlines associated with similar market directions closer"},{"finding":"The method can retrieve historically similar news days, e.g., matching a 2019‑11‑05 shutdown headline to a similar 2019‑05‑22 pipeline shutdown event.","kind":"other","quote":"our approach also identifies other days when similar events occurred. For instance, the most similar past headline might be “Russian Pipeline Shutdown Shifts Balance in Oil Market” (2019‑05‑22)."}],"reported_performance":null,"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":28,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:44:48Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}