No More Tuning: Prioritized Multi-Task Learning with Lagrangian Differential Multiplier Methods

Dan Ou; Qingwen Liu; Yuheng Huang; Zhengxing Cheng; Zhixuan Zhang

REVIEW 5 major objections 6 minor 29 references

No More Tuning: Prioritized Multi-Task Learning with Lagrangian Differential Multiplier Methods

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A Lagrange multiplier method gives multi-task models explicit priorities without loss-weight tuning.

desk verdict A simple Lagrangian wrapper with an interesting industrial A/B result, but the no-tuning claim is refuted by the paper's own appendix and the duality proof has a sign error, so it overclaims as written. read the letter →

arxiv 2412.12092 v1 pith:LT5NEOEF submitted 2024-12-16 cs.LG cs.IR

Zhengxing Cheng , Yuheng Huang , Zhixuan Zhang , Dan Ou , Qingwen Liu This is my paper · ORCID

classification cs.LGcs.IR

keywords multi-tasklearningtaskprioritizationconstrainedoptimizationLagrangianmultipliermethodsstrongdualityhyperparametertuningrecommendationsystemstorank

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that task priorities in multi-task learning can be enforced without manually balancing loss weights. Instead of weighting tasks, it solves a sequence of constrained problems: each lower-priority task is minimized subject to the constraint that all higher-priority tasks keep their previously achieved loss. The method, named NMT, uses a Lagrangian differential multiplier update (gradient descent on parameters, gradient ascent on multipliers) and the paper proves that, under a Lipschitz-smoothness and small-parameter-change assumption, strong duality holds, so the constrained problem can be solved as an unconstrained min-max problem. If correct, any gradient-based multi-task model can be given a priority ordering with no task-weight hyperparameters, and the primary task's performance is protected by construction. Experiments on two recommendation datasets and an online Taobao search A/B test report gains in the priority task and secondary tasks without the degradation seen with loss weighting.

What carries the argument

The mechanism is the Lagrangian differential multiplier method applied to each constrained subproblem. For stage k, the objective is L(θ, λ) = f_k(θ) + Σ_{j<k} λ_j (f_j(θ) − f_j(θ*)), with the constraint f_j(θ) ≤ f_j(θ*) for higher-priority tasks. Parameters θ are updated by gradient descent and multipliers λ by gradient ascent, giving an unconstrained min-max problem whose strong-duality guarantee (Theorem 1) rests on the perturbation function P(ξ) being approximately convex when the optimal parameters for neighboring tolerances are within ε. A re-scaling of the loss by 1/(1+Σλ_j) keeps the combined objective a normalized convex combination, avoiding loss explosion.

What would settle it

On a two-task problem where the secondary task has its global optimum far from the primary task's optimum in parameter space, run the NMT algorithm and check the constraint f1(θ) ≤ f1(θ*) at the final iterate. If it is violated, the strong-duality guarantee does not hold for that setting; a more direct check is to numerically evaluate the perturbation function P(ξ) near the optimum and test whether it is convex over the ξ range the optimizer actually visits.

Watch

Extended reading notes

Core claim

The central discovery claim is that a lexicographic-style prioritization in multi-task learning—minimize f1, then minimize f2 subject to f1 ≤ f1*, then f3 subject to f1 ≤ f1* and f2 ≤ f2*, and so on—can be carried out by a simple primal-dual gradient algorithm with no task-weight hyperparameters. The constrained problem (CO) and its Lagrangian dual (DO) are shown to have strong duality under two assumptions: no overfitting, and Lipschitz continuity of the losses together with bounded parameter changes inside the feasible region. Consequently, the dual ascent-descent updates converge to a solution that respects the priority constraints, which the authors verify by integrating NMT with Shared-Bottom, OMoE, MMoE, PLE, and FAMO on TikTok and QK-Video, and by deploying it in Taobao search with order volume as top priority, where it improved lower-priority metrics without sacrificing order volume.

Load-bearing premise

The guarantee that the primary task stays uncompromised assumes the optimal parameters for lower-priority tasks remain within a very small neighborhood of the primary task's optimum, where the loss functions are smooth and the constraint set is effectively convex.

Editorial extensions

If this is right

Any gradient-descent multi-task model can be given an explicit priority order by running NMT in stages; no task-balancing weights are introduced.
The tuning cost for m-task priorities drops from exponential grid search O(p^m) to m sequential optimization steps.
The primary task's loss is kept at its pre-optimized level (within the assumptions) while secondary tasks improve.
In the Taobao search deployment, adding lower-priority tasks under NMT improved their business metrics without the drop in order volume observed when the same tasks were added by loss weighting.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

"No tuning" should be read as "no task-balancing weights": the method still relies on hyperparameters such as learning rates for θ and λ, and the appendix reports a grid search over those; a fair test of the no-tuning claim would fix these values without search.
The strong-duality proof is local: it requires ε small, meaning the secondary task's best parameters must stay near the primary optimum; for tasks whose optima are far apart in parameter space, the constraint may be violated in practice even if the algorithm appears to converge.
The sequential scheme suggests a natural extension to safety-constrained learning: any constraint that can be expressed as an inequality on a loss can be enforced by the same multiplier update, potentially replacing reward shaping in constrained reinforcement learning.
An empirical check of the theory would be to compute the perturbation function P(ξ) on a small two-task network; if it is non-convex across the ξ range the optimizer visits, Theorem 1's premise is not satisfied.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The algorithm is simple and the industrial data point is interesting, but the two central promises—no tuning and a theoretical guarantee—are unsupported by the paper's own content.

What is new: the NMT wrapper, which first trains the primary task and then sequentially trains secondary tasks under a Lagrangian constraint, re-scaling the combined loss to avoid explosion. It is easy to implement and plugs into any gradient-descent MTL architecture. That is a legitimate, if modest, algorithmic contribution. The Taobao A/B test comparing pay+relevance against weight adjustment is a genuine industrial data point; the +0.26% order volume with +0.72% relevance, versus -0.35% order volume with weight adjustment, is worth paying attention to.

The soft spots are proportionally large. The proof of Proposition 2 in the appendix derives an absolute error bound (16) and then asserts a one-sided inequality (18). An absolute-value bound does not give you the sign, so the approximate convexity of the perturbation function is not established, and Theorem 1 collapses. Assumption 2 essentially assumes the parameters stay in a small epsilon-neighborhood—the conclusion—and Figure 3b shows the pay loss drifting above its optimum by up to 0.06 during training. So the 'primary task is preserved' guarantee is not delivered.

The 'No More Tuning' claim is directly contradicted by the appendix grid search over learning rates for both θ and λ. That is not a quibble; it is the main selling point. The experiments also lack error bars, and hyperparameters are selected on the evaluation datasets.

What remains is a heuristic that may work well in practice. The public-dataset gains are consistent but small, and without error bars they are hard to weigh. A serious referee could help the authors strip the overclaims and either fix the proof or present the method as a heuristic. I would send this to review and expect major revision, not desk reject, because the industrial A/B result and the simplicity of the method give the community something to test. But in its current form, I would not accept.

Referee Report

5 major / 6 minor

Summary. The manuscript proposes No More Tuning (NMT), a multi-task learning framework that formulates prioritized optimization as a sequence of constrained problems: the highest-priority task is minimized first, and each lower-priority task is then minimized subject to constraints that keep higher-priority losses at or below their previously obtained values. The constraints are handled by a Lagrangian, with gradient descent on parameters and gradient ascent on Lagrange multipliers. The paper claims that NMT requires no additional hyper-parameters, provides theoretical guarantees through a strong-duality result (Theorem 1) under Assumptions 1 and 2, and reports offline gains on TikTok and QK-Video and online gains in a Taobao search A/B test. The central promise is that secondary tasks can be improved while the primary task's performance is preserved, without manual loss-weight tuning.

Significance. If the theoretical claim and experiments were reliable, the paper would offer a practically attractive way to enforce priority in MTL without manual loss weighting, and the Taobao deployment gives useful evidence of applicability. The framework's integration with existing MTL architectures is a strength, and the A/B test reports positive business metrics for the primary task. However, the central theoretical argument is not established, the no-tuning claim is contradicted by the algorithm's own inputs and the appendix's grid search, and the experimental reporting lacks error bars. As it stands, the paper's main contribution is not supported.

major comments (5)

[Appendix, Proof of Proposition 2] The proof of Proposition 2 is invalid at the step from Eq. (16) to Eq. (18). Eq. (16) is an absolute-value bound |f_i(θ_t) - (t f_i(θ*(ξ1)) + (1-t) f_i(θ*(ξ2)))| ≤ Lε/2, but Eq. (18) asserts the one-sided inequality f_i(θ_t) ≤ t f_i(θ*(ξ1)) + (1-t) f_i(θ*(ξ2)) without establishing that the error term has the correct sign. An absolute bound does not rule out f_i(θ_t) exceeding the convex combination, and when the constraints at ξ1 and ξ2 are active there is no slack to absorb a positive error. Therefore the feasibility of θ_t is not proved, the approximate convexity of P(ξ) is not established, and Theorem 1, which rests on Proposition 2, is unsupported.
[Theoretical Analysis, Assumption 2 and Figure 3] Assumption 2 assumes that parameter differences inside the feasible region are bounded by an arbitrarily small ε and that this makes the perturbation function approximately convex, but the paper never connects ε to the optimization path of the deep networks used in the experiments. Figure 3b shows L_pay(θ) - L_pay(θ*) fluctuating up to about 0.06 above zero during training, i.e., the primary-task constraint is violated; the stated guarantee therefore does not apply to the reported training runs. The small-ε condition effectively assumes the conclusion that optimization stays within a region where the primary loss is preserved.
[Algorithm 1 and Appendix: Training Detail] The central claim that NMT requires 'no additional hyper-parameters' is contradicted by Algorithm 1's inputs: the parameter learning rate η, the multiplier learning rate τ, the initial multiplier λ_init, and per-stage convergence thresholds are all user-set quantities, and the tolerance r_i appears in the theoretical problem (CO). The appendix confirms this by describing a grid search over nine combinations of η and τ. At minimum the claim must be weakened to 'no task-weight hyper-parameters', and the sensitivity to these remaining hyper-parameters must be reported.
[Re-scaling method and Eq. (7)] The implemented algorithm uses the re-scaled loss L = 1/(1+Σ_j λ_j) (f_k(θ) + Σ_j λ_j (f_j(θ) - f_j(θ*_j))), while the theoretical analysis in Eq. (7) and Theorem 1 applies to the unscaled Lagrangian. The scaling factor is a function of λ and therefore changes the gradient dynamics of both θ and λ; the paper does not show that the strong-duality argument, even if valid, transfers to the re-scaled objective actually optimized in all experiments. This is a gap between theory and implementation.
[Tables 1–3 and Figure 2] The experimental claims are reported as point estimates without error bars, standard deviations, or significance tests, and the online A/B results are percentage deltas without confidence intervals. Given that many reported differences are on the order of 0.01%–0.5%, it is not possible to judge whether NMT improves the primary task or whether the differences are within run-to-run variation. This is particularly important for the paper's central claim that the primary task is never compromised.

minor comments (6)

[Introduction, bullet 'No Need for Parameter Adjustments'] The text says task prioritization is embedded in 'equality constraints', but the formulation throughout the paper uses inequality constraints; the terminology should be made consistent.
[Online Experiments, Eq. (11)] The notation L_pay(θ*) is used in Eq. (11) without defining θ* in the context of the online experiment; it should be clarified whether θ* is the single-task pay-optimized parameter vector.
[Theoretical Analysis, Assumption 1] Assumption 1 ('free from over-fitting') is not a formal mathematical condition; it should be replaced with a precise statement about the relationship between training loss and target performance.
[Throughout] There are typographical and formatting issues, including 'involxd' in the Introduction, 'Y uan' in the references, and a duplicated parenthesis in the statement of problem (CO).
[Main text and Appendix] The main text says the detailed proof of Proposition 2 will be provided in 'the Appendix of the extended version, which has been published on arXiv'; a journal or conference submission should be self-contained, and the proof should appear in the manuscript itself.
[Figure 2] The figure caption does not fully explain whether the colored lines are Pareto fronts or simple scans of two weights, nor whether the NMT points are obtained with the same computational budget; this should be clarified.

Circularity Check

2 steps flagged · score 6.0 of 10

The central strong-duality guarantee is circular: Proposition 2's proof assumes the one-sided convexity it must prove, and the 'primary task preserved' claim is the constraint itself.

other [Appendix, Proof of Proposition 2 (Eqs. 16–18), used by Theorem 1 in 'Theoretical Analysis']
"Since ǫ is proportional to the distance ‖θ∗ (ξ2)− θ∗ (ξ1)‖, and given that ǫ is small enough, O(ǫ) is negligible, and fi behaves approximately linearly in the convex region. ... Thus: fi(tθ∗ (ξ1) + (1−t)θ∗ (ξ2))≤ tfi(θ∗ (ξ1)) + (1−t)fi(θ∗ (ξ2)) ≤ t(fi(θ∗ ) + ri− ξ1i) + (1−t)(fi(θ∗ ) + ri− ξ2i)."

Eq. (16) is a two-sided absolute error bound: |fi(θt) − (t fi(θ*(ξ1)) + (1−t) fi(θ*(ξ2)))| ≤ Lε/2. The proof then asserts (17)–(18), dropping the error to obtain the one-sided inequality fi(θt) ≤ t fi(θ*(ξ1)) + (1−t) fi(θ*(ξ2)). That one-sided inequality is exactly the approximate convexity of the perturbation function P that Proposition 2 is supposed to prove. Theorem 1 then invokes 'convexity of the perturbation function' to assert strong duality and says 'we can get a feasible solution by optimizing the unconstrained dual problem DO.' Thus the feasibility/preservation guarantee is assumed in the small-ε 'approximately linear' premise rather than derived; no slack estimate is provided to absorb the error when constraints are active.
self definitional [Problem Formulation, Eq. (4); Algorithm 1, step 5, 'Compute the aggregate loss for task k']
"The optimization problem for f2(θ) is then: min θ f2(θ) s.t. f1(θ)≤ f1(θ∗ ) (4) ... Compute the aggregate loss for task k: L = fk(θ) + k−1∑ j=1 λj· (fj(θ)− fj(θ∗ ))."

The paper's central claim that NMT 'manages task prioritization ... by ensuring that secondary tasks are optimized without compromising the performance of the primary task' is not an independently derived prediction: the 'not compromised' condition is literally the constraint f1(θ) ≤ f1(θ*) used to define the surrogate problem, and the same term fj(θ)−fj(θ*) is inserted directly into the minimized Lagrangian. Any point satisfying the constraint is, by definition, one where the primary loss is no worse than θ*. The empirical plot Fig.3.b measures exactly this constraint violation.

full rationale

The paper is a genuine Lagrangian constrained-optimization method with external experiments on TikTok, QK-Video, and Taobao search, so not everything reduces to a self-citation chain; there are no load-bearing self-citations. However, the central theoretical claim that NMT 'ensures optimization guarantees' and preserves the primary task rests on Theorem 1, which depends on Proposition 2. The appendix proof of Proposition 2 derives only an absolute error bound and then silently replaces it with the one-sided inequality needed for convexity, effectively assuming the approximate convexity it must prove. Theorem 1 then uses that assumed convexity to conclude strong duality and feasibility of the dual solution. In parallel, the 'primary task is not compromised' headline is simply the constraint f1(θ) ≤ f1(θ*) embedded in the Lagrangian, so the empirical demonstration in Fig.3.b is a plot of the constraint violation rather than an independent test of a derived guarantee. This makes the theoretical guarantee partially circular by construction. Separately, the 'no tuning' claim is weakened by the grid search over the Lagrange-multiplier learning rate τ reported in the appendix, but that is a consistency concern rather than a circularity. Overall, the method may work in practice, but its stated first-principles guarantee is not established independently of its own assumptions.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The framework imports a classical Lagrangian solver and adds a re-scaling heuristic. The main uncharged inputs are the small-step duality assumption, the convergence of the alternating primal-dual updates, and several hyperparameters (learning rates, lambda_init, stage stopping rule) that the abstract claims do not exist.

free parameters (5)

theta learning rate eta = 1e-4, 5e-4, 1e-3 grid-searched
Grid search over nine combinations reported in the appendix; chosen on the same datasets used for evaluation, contradicting the no-tuning claim.
lambda learning rate tau = 1e-2, 5e-2, 1e-1 grid-searched
Same grid search as the theta learning rate, adding a second tuning dimension.
initial Lagrange multiplier lambda_init = not specified; Fig. 3 shows 0
The initial value of multipliers is an extra hyperparameter; the paper only says 'lambda_init' with no default or sensitivity analysis.
tolerance r_i in CO = not used in Algorithm 1
The theoretical relaxed problem introduces small positive tolerances r_i, but the actual algorithm uses r = 0; this mismatch is unresolved.
per-stage convergence threshold = not specified
The algorithm stops each stage 'until convergence' without defining a stopping rule, which affects the measured results.

assumptions (6)

standard math Convexity of f_i when strong duality is invoked for logistic regression
Used in the sentence 'If f_i is a convex function ... the strong duality between CO and DO holds (Platt and Barr 1987)'.
domain assumption Assumption 1: training is free from over-fitting
Stated to guarantee monotonicity between objective and performance; not verifiable from the experiments.
ad hoc to paper Assumption 2: theta differences bounded by epsilon and f_i Lipschitz
This is the load-bearing premise for Proposition 2 and Theorem 1; no bound connects epsilon to real training dynamics.
standard math Slater's condition holds because r_i > 0
Used to invoke strong duality; holds only in the relaxed problem with positive tolerance, not in the algorithm with r = 0.
standard math Rockafellar's perturbation-function theorem
Used in the proof of Theorem 1.
ad hoc to paper Gradient descent/ascent on L converges to a saddle point of DO
The paper assumes the simultaneous primal-dual updates solve the max-min problem, but no convergence proof for the alternating updates is given.

how reviews work

0 comments

Cite this review

Pith. "Pith review of No More Tuning: Prioritized Multi-Task Learning with Lagrangian Differential Multiplier Methods." pith.science (2026). https://pith.science/paper/LT5NEOEF

@misc{pith2026241212092,
  author       = {Pith},
  title        = {Pith review of: No More Tuning: Prioritized Multi-Task Learning with Lagrangian Differential Multiplier Methods},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LT5NEOEF}},
  note         = {Machine review of arXiv:2412.12092}
}

read the original abstract

Given the ubiquity of multi-task in practical systems, Multi-Task Learning (MTL) has found widespread application across diverse domains. In real-world scenarios, these tasks often have different priorities. For instance, In web search, relevance is often prioritized over other metrics, such as click-through rates or user engagement. Existing frameworks pay insufficient attention to the prioritization among different tasks, which typically adjust task-specific loss function weights to differentiate task priorities. However, this approach encounters challenges as the number of tasks grows, leading to exponential increases in hyper-parameter tuning complexity. Furthermore, the simultaneous optimization of multiple objectives can negatively impact the performance of high-priority tasks due to interference from lower-priority tasks. In this paper, we introduce a novel multi-task learning framework employing Lagrangian Differential Multiplier Methods for step-wise multi-task optimization. It is designed to boost the performance of high-priority tasks without interference from other tasks. Its primary advantage lies in its ability to automatically optimize multiple objectives without requiring balancing hyper-parameters for different tasks, thereby eliminating the need for manual tuning. Additionally, we provide theoretical analysis demonstrating that our method ensures optimization guarantees, enhancing the reliability of the process. We demonstrate its effectiveness through experiments on multiple public datasets and its application in Taobao search, a large-scale industrial search ranking system, resulting in significant improvements across various business metrics.

Figures

Figures reproduced from arXiv: 2412.12092 by the authors.

**Figure 2.** AUC performance comparison for different model [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Training metrics of secondary task relevance of pay (High priority) + relevance (Low priority). The top line shows the λ during training for the pay + relevance tasks. The middle line illustrates the fluctuation of the pay task loss around its optimal value. The bottom line displays the loss function of the secondary task relevance during the training process. finds a balance between the pay and relevance tasks, ens… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 16 canonical work pages

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Cao, Z.; Qin, T.; Liu, T.-Y.; Tsai, M.-F.; and Li, H. 2007. Learning to rank: from pairwise approach to listwise approach. In Proceedings of the 24th International Conference on Machine Learning, ICML '07, 129–136. New York, NY, USA: Association for Computing Machinery. ISBN 9781595937933

work page 2007
[4]

Carmel, D.; Haramaty, E.; Lazerson, A.; and Lewin-Eytan, L. 2020. Multi-Objective Ranking Optimization for Product Search Using Stochastic Label Aggregation. In Proceedings of The Web Conference 2020, WWW '20, 373–383. New York, NY, USA: Association for Computing Machinery. ISBN 9781450370233

work page 2020
[5]

Caruana, R. 1997. Multitask Learning. Machine Learning, 28: 41--75

work page 1997
[6]

Chen, S.; Zhang, Y.; and Yang, Q. 2024. Multi-task learning in natural language processing: An overview. ACM Computing Surveys, 56(12): 1--32

work page 2024
[7]

Chen, Z.; Badrinarayanan, V.; Lee, C.-Y.; and Rabinovich, A. 2018. GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks. arXiv:1711.02257

arXiv 2018
[8]

Hazimeh, H.; Zhao, Z.; Chowdhery, A.; Sathiamoorthy, M.; Chen, Y.; Mazumder, R.; Hong, L.; and Chi, E. H. 2021. DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning. arXiv:2106.03760

arXiv 2021

Show all 29 references

[9]

Javaloy, A.; and Valera, I. 2022. RotoGrad: Gradient Homogenization in Multitask Learning. arXiv:2103.02631

2022 arXiv
[10]

P.; and Ba, J

Kingma, D. P.; and Ba, J. 2017. Adam: A Method for Stochastic Optimization. arXiv:1412.6980

2017 arXiv
[11]

Kokkinos, I. 2016. UberNet: Training a `Universal' Convolutional Neural Network for Low-, Mid-, and High-Level Vision using Diverse Datasets and Limited Memory. arXiv:1609.02132

2016 arXiv
[12]

Lin, X.; Chen, H.; Pei, C.; Sun, F.; Xiao, X.; Sun, H.; Zhang, Y.; Ou, W.; and Jiang, P. 2019. A pareto-efficient algorithm for multiple objective optimization in e-commerce recommendation. In Proceedings of the 13th ACM Conference on Recommender Systems, RecSys '19, 20–28. Ne...

2019
[13]

Liu, B.; Feng, Y.; Stone, P.; and Liu, Q. 2024. Famo: Fast adaptive multitask optimization. Advances in Neural Information Processing Systems, 36

2024
[14]

Ma, J.; Zhao, Z.; Yi, X.; Chen, J.; Hong, L.; and Chi, E. H. 2018. Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts. KDD '18, 1930–1939. New York, NY, USA: Association for Computing Machinery. ISBN 9781450355520

2018
[15]

Mahapatra, D.; Dong, C.; Chen, Y.; Meng, D.; and Momma, M. 2022. Multi-Label Learning to Rank through Multi-Objective Optimization. arXiv:2207.03060

2022 arXiv
[16]

Misra, I.; Shrivastava, A.; Gupta, A.; and Hebert, M. 2016. Cross-stitch Networks for Multi-task Learning. arXiv:1604.03539

2016 arXiv
[17]

Paternain, S.; Chamon, L. F. O.; Calvo-Fullana, M.; and Ribeiro, A. 2019. Constrained Reinforcement Learning Has Zero Duality Gap. arXiv:1910.13393

2019 arXiv
[18]

Platt, J.; and Barr, A. 1987. Constrained Differential Optimization. In Anderson, D., ed., Neural Information Processing Systems, volume 0. American Institute of Physics

1987
[19]

Rockafellar, R. T. 1970. Convex analysis. Princeton Mathematical Series. Princeton, N. J.: Princeton University Press

1970
[20]

Sermanet, P.; Eigen, D.; Zhang, X.; Mathieu, M.; Fergus, R.; and LeCun, Y. 2014. OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks. arXiv:1312.6229

2014 arXiv
[21]

Stooke, A.; Achiam, J.; and Abbeel, P. 2020. Responsive Safety in Reinforcement Learning by PID Lagrangian Methods. arXiv:2007.03964

2020 arXiv
[22]

Su, L.; Pan, J.; Wang, X.; Xiao, X.; Quan, S.; Chen, X.; and Jiang, J. 2024. STEM: Unleashing the Power of Embeddings for Multi-task Recommendation. arXiv:2308.13537

2024 arXiv
[23]

Tang, H.; Liu, J.; Zhao, M.; and Gong, X. 2020. Progressive Layered Extraction (PLE): A Novel Multi-Task Learning (MTL) Model for Personalized Recommendations. In Proceedings of the 14th ACM Conference on Recommender Systems, RecSys '20, 269–278. New York, NY, USA: Association...

2020
[24]

Tang, J.; Gao, H.; He, L.; and Katariya, S. 2024. Multi-objective Learning to Rank by Model Distillation. arXiv preprint arXiv:2407.07181

2024 arXiv
[25]

J.; and Mannor, S

Tessler, C.; Mankowitz, D. J.; and Mannor, S. 2018. Reward Constrained Policy Optimization. arXiv:1805.11074

2018 arXiv
[26]

T.; Wong, Y.; Liu, Z.; Zhao, X.; Wang, Y.; Chen, B.; Guo, H.; and Tang, R

Wang, Y.; Lam, H. T.; Wong, Y.; Liu, Z.; Zhao, X.; Wang, Y.; Chen, B.; Guo, H.; and Tang, R. 2023. Multi-task deep recommender systems: A survey. arXiv preprint arXiv:2302.03525

2023 arXiv
[27]

Woo, H.; Lee, H.; and Cho, S. 2021. An Efficient Combinatorial Optimization Model Using Learning-to-Rank Distillation. arXiv:2201.00695

2021 arXiv
[28]

Yu, T.; Kumar, S.; Gupta, A.; Levine, S.; Hausman, K.; and Finn, C. 2020. Gradient Surgery for Multi-Task Learning. arXiv:2001.06782

2020 arXiv
[29]

Yuan, G.; Yuan, F.; Li, Y.; Kong, B.; Li, S.; Chen, L.; Yang, M.; Yu, C.; Hu, B.; Li, Z.; Xu, Y.; and Qie, X. 2023. Tenrec: A Large-scale Multipurpose Benchmark Dataset for Recommender Systems. arXiv:2210.10629

2023 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Cao, Z.; Qin, T.; Liu, T.-Y.; Tsai, M.-F.; and Li, H. 2007. Learning to rank: from pairwise approach to listwise approach. In Proceedings of the 24th International Conference on Machine Learning, ICML '07, 129–136. New York, NY, USA: Association for Computing Machinery. ISBN 9781595937933

work page 2007

[4] [4]

Carmel, D.; Haramaty, E.; Lazerson, A.; and Lewin-Eytan, L. 2020. Multi-Objective Ranking Optimization for Product Search Using Stochastic Label Aggregation. In Proceedings of The Web Conference 2020, WWW '20, 373–383. New York, NY, USA: Association for Computing Machinery. ISBN 9781450370233

work page 2020

[5] [5]

Caruana, R. 1997. Multitask Learning. Machine Learning, 28: 41--75

work page 1997

[6] [6]

Chen, S.; Zhang, Y.; and Yang, Q. 2024. Multi-task learning in natural language processing: An overview. ACM Computing Surveys, 56(12): 1--32

work page 2024

[7] [7]

Chen, Z.; Badrinarayanan, V.; Lee, C.-Y.; and Rabinovich, A. 2018. GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks. arXiv:1711.02257

arXiv 2018

[8] [8]

Hazimeh, H.; Zhao, Z.; Chowdhery, A.; Sathiamoorthy, M.; Chen, Y.; Mazumder, R.; Hong, L.; and Chi, E. H. 2021. DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning. arXiv:2106.03760

arXiv 2021