{"paper_id":"2609.39261","title":"Jacobian Rank Collapse in Decision-Focused Learning","authors":["Aojie Yuan","Haiyue Zhang","Zijian Su"],"paper_date":"2026-09-30","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2609.39261","page_url":"https://thequant.space/flowcharts/jacobian-rank-collapse-in-decision-focused-learning/","scores":{"empirical_rigor":8,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes to characterize decision-focused learning restrictions via the predictor Jacobian and to study how Jacobian rank affects learning directions and decision quality.","quote":"We characterize this restriction through the predictor Jacobian, using sparse index tracking to distinguish the covariance entries read by the optimizer from the parameter directions available to learning.","summary":"Decision-focused learning (DFL) trains predictors through downstream objectives, but a different loss need not provide an independent parameter-update direction. We characterize this restriction through the predictor Jacobian, using sparse index tracking to distinguish the covariance entries read…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Isotropy of the normalized energy of gradient coordinates for a fixed support.","type":"statistical","stated":"explicit","quote":"For a ﬁxed support, a sufﬁcient assumption is isotropy of the normalized energy: E[g2ij/∥g∥2] = 1/N2 for every coordinate."},{"assumption":"Hard‑K cardinality constraint (exact K assets) is used instead of ℓ1 relaxation.","type":"model","stated":"explicit","quote":"All experiments therefore use hard-K."},{"assumption":"Dynamic stock selection changes the selection rule and can hurt performance.","type":"execution","stated":"explicit","quote":"Dynamic selection (re‑selecting stocks at each rebalance by tracking score) combined with DFL hurts performance: TE = 0.0346 (vs. 0.0287 for ﬁxed) with turnover 6× higher."},{"assumption":"Invertible coordinate scaling changes SGD outcomes without changing expressivity.","type":"model","stated":"explicit","quote":"Holding expressivity fixed, invertible coordinate scaling lowers spectral effective rank and ordinary SGD gains; compensating for the scaling restores the original trajectories."},{"assumption":"Transaction‑cost penalty is linear in the ℓ1 norm of weight changes.","type":"execution","stated":"explicit","quote":"We augment the QP with an explicit transaction cost penalty: minw w⊤ˆΣw −2w⊤σidx + c∥w − wprev∥1"}]},"datasets":{"status":"available","items":[{"name":"38 one‑parameter equity configurations","provider":null,"kind":"market-prices","access":"public","period":null,"universe":"Various markets (S&P 500, FTSE 100, Nikkei 225, Euro Stoxx 50, ASX 200, Hang Seng, S&P 150","quote":"Across 38 one‑parameter equity configurations, DFL gains over MSE remain below 1.8%; a 385‑parameter conditional predictor also has pointwise rank one."},{"name":"Shortest‑path synthetic datasets","provider":null,"kind":"synthetic","access":"public","period":null,"universe":null,"quote":"Each task uses ten independently generated datasets, a shared train‑only ridge initializer, and identical validation budgets for MSE and SPO+ at every capacity."},{"name":"Knapsack synthetic datasets","provider":null,"kind":"synthetic","access":"public","period":null,"universe":null,"quote":"Each task uses ten independently generated datasets, a shared train‑only ridge initializer, and identical validation budgets for MSE and SPO+ at every capacity."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":null,"detail":"Cost coefficient c is set to 100 bps; turnover is 0.030–0.046 vs. 0.002–0.003, adding only 4 bps net cost at K=10 and 3 bps at K=20.","quote":"... even at c = 100 bps it adds only 4 bps of net cost at K=10 and 3 bps at K=20, against a TE reduction of 49 and 57 bps respectively.","source":"evidence_card","rule":"R2","candidates":{"front_matter":"unclear","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Cost coefficient c is set to 100 bps; turnover is 0.030–0.046 vs. 0.002–0.003, adding only 4 bps net cost at K=10 and 3 bps at K=20.","card_quote":"... even at c = 100 bps it adds only 4 bps of net cost at K=10 and 3 bps at K=20, against a TE reduction of 49 and 57 bps respectively.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"DFL gains over MSE are modest, staying below 1.8% across 38 equity configurations.","kind":"performance","quote":"Across 38 one‑parameter equity configurations, DFL gains over MSE remain below 1.8%; a 385‑parameter conditional predictor also has pointwise rank one."},{"finding":"Full‑capacity SPO+ reduces mean test regret by 11.59% on shortest‑path and 10.59% on knapsack, with only knapsack passing Holm correction across eight tests.","kind":"performance","quote":"Full‑capacity SPO+ reduces mean test regret by 11.59% on shortest path and 10.59% on knapsack."},{"finding":"Invertible coordinate scaling lowers spectral effective rank and harms ordinary SGD gains; compensating restores original trajectories.","kind":"theoretical","quote":"Holding expressivity fixed, invertible coordinate scaling lowers spectral effective rank and ordinary SGD gains; compensating for the scaling restores the original trajectories."},{"finding":"Matched neural forward‑target control finds no aggregate DFL advantage in the tested architecture.","kind":"negative","quote":"neural comparison has no aggregate DFL gain in its tested architecture; this is a descriptive result, not an equivalence test."},{"finding":"Dynamic selection with BD‑DFL regularization can reduce tracking‑error relative to unregularized DFL, but gains are not universal.","kind":"performance","quote":"With dynamic tracking‑score selection at N=100, neural DFL increases mean TE by 8.6% at K=10; BD‑DFL with β = 0.001 reduces it by 8.2% relative to MSE."}],"reported_performance":null,"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":39,"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":"2026-11-14T17:04:06Z"},"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/"}