betaFeedback

WorksComputer Science

From Task Affinities to Optimal Allocations: Probabilistic Graphical Modelling for Supervised Finetuning

Abstract

Supervised fine-tuning over a large pool of heterogeneous tasks is often computationally expensive, leading practitioners to train language models on a fixed training budget of carefully chosen SFT tasks. Thus, achieving strong downstream performance hinges on selecting an effective subset of tasks, or equivalently, identifying an optimal task mixture under a fixed budget constraint. In practice, supervised finetuning mixtures are often determined using simple heuristics (e.g., uniform or size-proportional sampling) that ignore task interactions, which can hurt transfer and waste budget on redundant task sources. We introduce TASKMIXPGM, a framework for learning continuous task mixtures through energy-based modelling over SFT tasks. Each task forms a node in a Markov random field, with unary potentials capturing task utility and pairwise potentials modelling inter-task relationships via behavioral divergences between predictive distributions of single-task fine-tuned models. Optimizing this objective yields task mixtures that balance utility, coverage, and redundancy. We further show that the induced set function defined over the set of such tasks is weakly submodular under budget constraints, enabling approximation guarantees for discrete task selection. Empirically, we provide a fixed, benchmark-independent mixture that improves over heuristic mixing across multiple model families and different training budgets. TASKMIXPGM also uncovers interpretable structure in task interactions, revealing complementarities and redundancies that inform effective SFT mixtures.

Introduction

Supervised fine-tuning (SFT) on a mixture of instruction datasets shapes much of a language model’s reasoning and problem-solving ability. With a fixed training budget and pools of hundreds of tasks, the central design choice is which tasks to include and how much of the budget each receives. Common practice relies on uniform sampling, dataset-size weighting [8], or hand-crafted rules, which can let dominant sources take over the mixture and spend budget on redundant tasks. Automated methods such as submodular selection (SMART; RBR24), influence-based retrieval (LESS, XMG+24; BIDS, DZMP25) and proxy-based mixture prediction (RegMix, LZM+24; Data Mixing Laws, LLX25) typically select discrete subsets rather than continuous proportions, rely on semantic or per-example scores that do not capture cross-task transfer, or require repeated training runs.

How can we automatically choose an SFT task mixture that is simultaneously representative and diverse, and that reliably improves downstream performance under a fixed budget?

We introduce TaskMixPGM (Task Mixtures via Probabilistic Graphical Modelling), which models the task pool as a dense Markov random field: unary potentials reward representative tasks and pairwise potentials penalize redundant ones. Both are computed from behavioral similarity, by comparing the predictions of single-task fine-tuned models, rather than from metadata or embeddings. The learned weights also show which tasks act as broad “attractors” and which are specialized, and for budgeted selection we show that the induced set function is weakly submodular, so greedy construction comes with approximation guarantees.

Q1. Can functional, behavior-based task affinities be used to compute continuous mixture proportions that outperform strong heuristic and selection baselines?

Q2. Does the resulting mixture optimization yield interpretable structure (task clusters, redundancy, and influential “attractor” tasks) and support efficient budgeted construction?

Contributions. Our main contributions are as follows:

  • An energy model over behavioral task affinities. We cast SFT mixture design as energy minimization over tasks, with affinities measured from how single-task models predict each other’s data.

  • Guarantees for budgeted task selection. We show that the induced set function is monotone and weakly submodular, so greedy task selection has an approximation guarantee.

  • Insight into what drives the mixture. We characterize how much spectral shift convexity actually requires, and show that the mixture favors broadly transferable tasks.

  • A single, benchmark-independent mixture. Computed once, the same mixture outperforms heuristic and selection baselines across three model families across different budgets on state-of-the-art benchmarks.

Preliminaries and Background work

Notation. We write [n]:={1,…,n}[n] := \{1,\ldots,n\} and R+n\mathbb{R}_{+}^{n} for the non-negative orthant. Vectors are lower-case bold (x\boldsymbol{x}) and matrices upper-case bold (X\boldsymbol{X}), with individual entries denoted as x[i]x_{[i]} and X[ij]X_{[ij]}. x[i:j]x_{[i:j]} denotes the subvector with indices i,…,j−1i,\ldots,j-1, and x[:j]:=x[0:j]x_{[:j]} := x_{[0:j]}. 1n\mathbf{1}_{n}, 1m×n\mathbf{1}_{m\times n} and I\mathbf{I} are the all-ones vector, the all-ones matrix and the identity. Δn:={x∈R+n:∑ix[i]=1}\Delta_{n} := \{x \in\mathbb{R}_{+}^{n} : \sum_i x_{[i]} = 1\} denotes the probability simplex, and (a)+:=max⁡{a,0}(a)_{+} := \max\{a,0\}. For μ,ν∈Δn\mu,\nu\in\Delta_n, KL⁡(μ∥ν):=∑iμ[i]log⁡(μ[i]/ν[i])\operatorname{KL}(\mu\parallel\nu) := \sum_i \mu_{[i]} \log(\mu_{[i]}/\nu_{[i]}) denotes the Kullback–Leibler divergence, assuming ν[i]>0\nu_{[i]} > 0 whenever μ[i]>0\mu_{[i]} > 0. For symmetric matrix X\boldsymbol{X}, λmin⁡(X)\lambda_{\min}(\boldsymbol{X}) and λmax⁡(X)\lambda_{\max}(\boldsymbol{X}) are its extreme eigenvalues, and for X≻0\boldsymbol{X} \succ0, κ(X):=λmax⁡(X)/λmin⁡(X)\kappa(\boldsymbol{X}) := \lambda_{\max}(\boldsymbol{X})/\lambda_{\min}(\boldsymbol{X}) is its condition number. We write X⪰0\boldsymbol{X} \succeq0 for a positive semi-definite (PSD) matrix and X⋡0\boldsymbol{X} \nsucceq0 else.

Further, let 1=(1,…,1)⊤∈Rn\mathbf{1} = (1,\ldots,1)^{\top} \in\mathbb{R}^{n} denote the all-ones vector, and 1⊥={h∈Rn:1⊤h=0}\mathbf{1}^{\perp} = \{h \in\mathbb{R}^{n} : \mathbf{1}^{\top}h = 0\} be its orthogonal complement: the vectors whose entries sum to zero. If p,p′∈Δnp,p' \in\Delta_n, then 1⊤(p′−p)=1−1=0\mathbf{1}^{\top}(p' - p) = 1 - 1 = 0, so p′−p∈1⊥p' - p \in\mathbf{1}^{\perp}.

Submodularity ratio. Let f:2V→Rf : 2^{\mathcal{V}} \to\mathbb{R} be a set function over ground set V\mathcal{V}. For any subset S⊆VS \subseteq\mathcal{V} and element e∈V∖Se \in\mathcal{V} \setminus S, the marginal gain is Δf(e∣S)=f(S∪{e})−f(S)\Delta f(e \mid S) = f(S \cup\{e\}) - f(S). The function ff is submodular if Δf(e∣A)≥Δf(e∣B)\Delta f(e \mid A) \geq\Delta f(e \mid B) for all A⊆B⊆VA \subseteq B \subseteq\mathcal{V} and e∉Be \notin B. The submodularity ratio [12] captures how close a non-submodular function is to being submodular:

γL,κ(f)=min⁡S⊆L, A⊆V∣A∣≤κ, A∩S=∅∑u∈AΔf(u∣S)f(S∪A)−f(S).(1)\gamma_{\mathcal{L},\kappa}(f) = \min_{\substack{S \subseteq\mathcal{L},\ A \subseteq\mathcal{V}\\ |A|\leq\kappa,\ A\cap S=\varnothing}} \frac{\sum_{u\in A}\Delta f(u\mid S)}{f(S\cup A)-f(S)}. \tag*{(1)}

with 0/0:=10/0 := 1. If γ>0\gamma> 0, the greedy algorithm for max⁡∣S∣≤κf(S)\max_{|S|\leq\kappa} f(S) achieves a (1−e−γ)(1-e^{-\gamma}) approximation to the optimum [31, 12].

RSC and RSM. A differentiable function l:Rd→Rl : \mathbb{R}^{d} \to\mathbb{R} is restricted strong concave (RSC) with parameter cΩc_{\Omega} and restricted smooth (RSM) with parameter CΩC_{\Omega} on a domain Ω⊆Rd×Rd\Omega\subseteq\mathbb{R}^{d}\times\mathbb{R}^{d} if ∀(x,y)∈Ω\forall(\boldsymbol{x},\boldsymbol{y})\in\Omega,

l(x)−l(y)+⟨∇l(x),y−x⟩∈[cΩ2∥y−x∥22,CΩ2∥y−x∥22].l(\boldsymbol{x}) - l(\boldsymbol{y}) + \langle\nabla l(\boldsymbol{x}),\boldsymbol{y}-\boldsymbol{x}\rangle\in\left[\frac{c_{\Omega}}{2}\|\boldsymbol{y}-\boldsymbol{x}\|_{2}^{2},\frac{C_{\Omega}}{2}\|\boldsymbol{y}-\boldsymbol{x}\|_{2}^{2}\right].

RSC and RSM have been studied in the context of weakly submodular functions [15, 17]. Let Ωκ:={(x,y):∥x∥0≤κ, ∥y∥0≤κ, x,y≥0}\Omega_{\kappa} := \{(\boldsymbol{x},\boldsymbol{y}) : \|\boldsymbol{x}\|_{0} \leq\kappa,\ \|\boldsymbol{y}\|_{0} \leq\kappa,\ \boldsymbol{x},\boldsymbol{y} \geq0\}, and let cκc_{\kappa} and CκC_{\kappa} denote the RSC and RSM parameters of ll on Ωκ\Omega_{\kappa}. Since Ωk⊆Ωκ\Omega_k \subseteq\Omega_{\kappa} whenever k≤κk \leq\kappa, it is easily verified that ck≥cκc_k \geq c_{\kappa} and Ck≤CκC_k \leq C_{\kappa}. Finally, define Ω^:={(x,y):∥x−y∥0≤1}\widehat{\Omega}:=\{(x,y):\|x-y\|_0\leq1\}, with associated smoothness parameter C~1\widetilde{C}_1.

Related Work. Choosing the right mixture of training data is central to efficient LLM fine-tuning, whether the goal is a specific task or broad generalization. Early work measured task similarity with Maximum Mean Discrepancy or reconstruction error to retrieve examples [2, 19, 3]; more recent work represents tasks by lightweight LoRA adapters and uses low-rank similarity to guide transfer [22].

A second line of work estimates the influence of individual examples on a target task. Influence functions [20] are accurate but expensive, so cheaper proxies track forgotten examples [41] or gradient norms [PGD21]. For instruction tuning, LESS [49] retrieves examples whose low-rank gradient features match the target, and BIDS [14] normalizes these scores per task to avoid favoring high-gradient tasks.

Other work targets diversity and coverage, combining difficulty scores with clustering [52, 30] or selecting coresets that approximate full training dynamics [36, 24]. For multi-task instruction tuning, SMART [35] allocates the budget across tasks with a submodular objective and selects non-redundant examples. Because scoring every example on a large model is costly, STAFF [53] scores with a smaller sibling model first, TSDS [26] uses optimal transport for scalable distribution matching, and DELIFT [1] scores examples by their usefulness as in-context prompts for one another. See Section H for additional related work on data mixture optimisation and task-affinity-based reweighting.

Contextualizing our framework. These methods rank data by scalar relevance scores, either per instance [49, 14] or per dataset [2, 22], and do not model redundancy between tasks. We instead pose mixture selection as optimization over an energy landscape of pairwise task interactions, encoded in a symmetric affinity matrix SS estimated from the predictive behavior of single-task models.

Proposed Approach

Problem setup. We consider a set of nn instruction tuning tasks V:={T1,…,Tn}\mathcal{V}:=\{\mathcal{T}_1,\ldots,\mathcal{T}_n\}, where each task Ti\mathcal{T}_i is associated with a corresponding dataset Di\mathcal{D}_i. Typically we consider supervised finetuning instruction dataset of the form Di={xi,yi}i=1n\mathcal{D}_i=\{x_i,y_i\}_{i=1}^{n} where each xix_i is an input sequence and yiy_i is the corresponding output.

Our goal is to learn a sampling distribution p∗∈Δn\mathbf{p}^* \in\Delta^n over nn such tasks, which determines how training effort can be distributed across the collection of tasks.

Redundancy and structure across tasks. Tasks in a training pool are neither independent nor equally informative. Redundancy can arise from overlapping datasets across collections, such as BoolQ, RTE, and COPA appearing in both FLAN [25] and SuperGLUE [45], from multiple prompt variants of the same dataset in T0 [39], or from tasks targeting similar capabilities, such as MNLI, SNLI, and ANLI for natural language inference. In contrast, tasks such as Evol-CodeAlpaca for code generation, NuminaMath-TIR for mathematical reasoning, and Aya [SVD+24] for multilingual instruction following provide complementary coverage. Existing mixture methods often optimize tasks within predefined groups or relative to a fixed target [35, 26], which can miss these interactions. A task that is redundant in one mixture may be useful in another because of the other tasks selected. This motivates modeling the task pool jointly to account for redundancy, complementarity, and cross-task effects when allocating the training budget.

Cross-task Similarity. We collect pairwise task affinities in a symmetric matrix S∈Rn×nS \in\mathbb{R}^{n \times n}, where S[ij]S_{[ij]} measures how related Ti\mathcal{T}_i and Tj\mathcal{T}_j are (defined in Equation (5)). The degree S[i]:=∑jS[ij]S_{[i]}:=\sum_j S_{[ij]} summarizes how strongly Ti\mathcal{T}_i relates to the pool as a whole: tasks with larger degree share more structure with other tasks, while tasks with smaller degree are more specialized.

References

  1. [1] [AKPD25] Ishika Agarwal, Krishnateja Killamsetty, Lucian Popa, and Marina Danilevsky. DELIFT: Data efficient language model instruction fine-tuning. In The Thirteenth International Conference on Learning Representations, 2025.arxiv.org/abs/2411.04425
  2. [2] [ALT+19] Alessandro Achille, Michael Lam, Rahul Tewari, Avinash Ravichandran, Subhransu Maji, Charles C Fowlkes, Stefano Soatto, and Pietro Perona. Task2vec: Task embedding for meta-learning. In Proceedings of the IEEE/CVF international conference on computer vision, pages 6430–6439, 2019.arxiv.org/abs/1902.03545
  3. [3] [AMF20] David Alvarez-Melis and Nicolo Fusi. Geometric dataset distances via optimal transport. Advances in Neural Information Processing Systems, 33:21428–21439, 2020.
  4. [4] [BT03] Amir Beck and Marc Teboulle. Mirror descent and nonlinear projected subgradient methods for convex optimization. Operations Research Letters, 31(3):167–175, 2003.DOI
  5. [5] [BT09] Amir Beck and Marc Teboulle. A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIAM journal on imaging sciences, 2(1):183–202, 2009.DOI
  6. [6] [CGR09] Yihua Chen, Maya R Gupta, and Benjamin Recht. Learning kernels from indefinite similarities. In Proceedings of the 26th Annual International Conference on Machine Learning, pages 145–152, 2009.DOI
  7. [7] [CH90] Kenneth Ward Church and Patrick Hanks. Word association norms, mutual information, and lexicography. Computational Linguistics, 16(1):22–29, 1990.
  8. [8] [CHL+24] Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al. Scaling instruction-finetuned language models. Journal of Machine Learning Research, 25(70):1–53, 2024.
  9. [10] [CRB+23] Mayee F. Chen, Nicholas Roberts, Kush Bhatia, Jue Wang, Ce Zhang, Frederic Sala, and Christopher Ré. Skill-it! A data-driven skills framework for understanding and training language models. In Advances in Neural Information Processing Systems, 2023.arxiv.org/abs/2307.14430
  10. [11] [DK11] Abhimanyu Das and David Kempe. Submodular meets spectral: Greedy algorithms for subset selection, sparse approximation and dictionary selection. In Proceedings of the 28th International Conference on Machine Learning (ICML), pages 1057–1064, 2011.arxiv.org/abs/1102.3975
  11. [12] [DK18] A. Das and D. Kempe. Approximate submodularity and its applications: Subset selection, sparse approximation and dictionary selection. Journal of Machine Learning Research, 19(3):1–34, 2018.
  12. [14] [DZMP25] Qirun Dai, Dylan Zhang, Jiaqi W Ma, and Hao Peng. Improving influence-based instruction tuning data selection for balanced learning of diverse capabilities. arXiv preprint arXiv:2501.12147, 2025.arxiv.org/abs/2501.12147
  13. [15]Ethan R. Elenberg, Rajiv Khanna, Alexandros G. Dimakis, and Sahand Negahban. Restricted strong convexity implies weak submodularity. The Annals of Statistics, 46(6B):3539–3568, 2018.arxiv.org/abs/1612.00804
  14. [16]Christopher Fifty, Ehsan Amid, Zhe Zhao, Tianhe Yu, Rohan Anil, and Chelsea Finn. Efficiently identifying task groupings for multi-task learning. In Advances in Neural Information Processing Systems, 2021.arxiv.org/abs/2109.04617
  15. [17]Karthik S. Gurumoorthy, Amit Dhurandhar, Guillermo Cecchi, and Charu Aggarwal. Efficient data representation by selecting prototypes with importance weights. In IEEE International Conference on Data Mining (ICDM), pages 260–269, 2019. arXiv:1707.01212.arxiv.org/abs/1707.01212
  16. [18]Roger A. Horn and Charles R. Johnson. Matrix analysis. Cambridge university press, 2012.
  17. [19]MyungGwon Hwang, Yuna Jeong, and Wonkyung Sung. Data distribution search to select core-set for machine learning. In The 9th International conference on smart media and applications, pages 172–176, 2020.DOI
  18. [20]Pang Wei Koh and Percy Liang. Understanding black-box predictions via influence functions. In International conference on machine learning, pages 1885–1894. PMLR, 2017.arxiv.org/abs/1703.04730
  19. [21]Ross Kindermann and J Laurie Snell. Markov random fields and their applications, volume 1. American Mathematical Society, 1980.
  20. [22]Hw ichan Kim, Shota Sasaki, Sho Hoshino, and Ukyo Honda. A single linear layer yields task-adapted low-rank matrices. In Nicoletta Calzolari, Min-Yen Kan, Veronique Hoste, Alessandro Lenci, Sakriani Sakti, and Nianwen Xue, editors, Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 1602–1608, Torino, Italia, May 2024. ELRA and ICCL.arxiv.org/abs/2403.14946
  21. [23]Alex Kulesza and Ben Taskar. Determinantal point processes for machine learning. Foundations and Trends in Machine Learning, 5(2–3):123–286, 2012.
  22. [24]Krishnateja Killamsetty, Xujiang Zhao, Feng Chen, and Rishabh Iyer. Retrieve: Coreset selection for efficient and robust semi-supervised learning. Advances in neural information processing systems, 34:14488–14501, 2021.arxiv.org/abs/2106.07760
  23. [25]Shayne Longpre, Le Hou, Tu Vu, Albert Webson, Hyung Won Chung, Yi Tay, Denny Zhou, Quoc V Le, Barret Zoph, Jason Wei, et al. The flan collection: Designing data and methods for effective instruction tuning. In International conference on machine learning, pages 22631–22648. PMLR, 2023.arxiv.org/abs/2301.13688
  24. [26]Zifan Liu, Amin Karbasi, and Theodoros Rekatsinas. TSDS: Data selection for task-specific model finetuning. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024.arxiv.org/abs/2410.11303
  25. [28]Nathan Lambert, Jacob Morrison, Valentina Pyatkin, Shengyi Huang, Hamish Ivison, Faeze Brahman, Lester James V Miranda, Alisa Liu, Nouha Dziri, Shane Lyu, et al. Tulu 3: Pushing frontiers in open language model post-training. arXiv preprint arXiv:2411.15124, 2024.arxiv.org/abs/2411.15124
  26. [29]Qian Liu, Xiaosen Zheng, Niklas Muennighoff, Guangtao Zeng, Longxu Dou, Tianyu Pang, Jing Jiang, and Min Lin. Regmix: Data mixture as regression for language model pre-training. arXiv preprint arXiv:2407.01492, 2024.arxiv.org/abs/2407.01492
  27. [30]Adyasha Maharana, Prateek Yadav, and Mohit Bansal. D2 pruning: Message passing for balancing diversity and difficulty in data pruning. arXiv preprint arXiv:2310.07931, 2023.arxiv.org/abs/2310.07931
  28. [31]G.L. Nemhauser, L.A. Wolsey, and M.L. Fisher. An analysis of approximations for maximizing submodular set functions—i. Mathematical Programming, 14:265–294, 1978.DOI
  29. [32]Guillermo Ortiz-Jimenez, Alessandro Favero, and Pascal Frossard. Task arithmetic in the tangent space: Improved editing of pre-trained models. arXiv preprint arXiv:2305.12827, 2023.
  30. [34]Qwen Team. Qwen3.5: Towards native multimodal agents, February 2026.
  31. [35]H S V N S Kowndinya Renduchintala, Sumit Bhatia, and Ganesh Ramakrishnan. SMART: Submodular data mixture strategy for instruction tuning. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar, editors, Findings of the Association for Computational Linguistics: ACL 2024, pages 12916–12934, Bangkok, Thailand, August 2024. Association for Computational Linguistics.
  32. [36]Ozan Sener and Silvio Savarese. Active learning for convolutional neural networks: A core-set approach. In International Conference on Learning Representations, 2018.
  33. [37]Lintang Sutawika, Hailey Schoelkopf, Leo Gao, Baber Abbasi, Stella Biderman, Jonathan Tow, Charles Lovering, Jason Phang, Anish Thite, Thomas Wang, et al. Eleutherai/lm-evaluation-harness: v0. 4.9. Zenodo, 2024.
  34. [39]Victor Sanh, Albert Webson, Colin Raffel, Stephen H Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Teven Le Scao, Arun Raja, et al. Multitask prompted training enables zero-shot task generalization. arXiv preprint arXiv:2110.08207, 2021.arxiv.org/abs/2110.08207
  35. [40]Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023.arxiv.org/abs/2307.09288
  36. [41]Mariya Toneva, Alessandro Sordoni, Remi Tachet des Combes, Adam Trischler, Yoshua Bengio, and Geoffrey J. Gordon. An empirical study of example forgetting during deep neural network learning. In International Conference on Learning Representations, 2019.arxiv.org/abs/1812.05159
  37. [42]Gang Wu, Edward Y Chang, and Zhihua Zhang. An analysis of transformation on non-positive semidefinite similarity matrix for kernel machines. In Proceedings of the 22nd international conference on machine learning, volume 8. Citeseer Cham, 2005.
  38. [43]Martin J. Wainwright and Michael I. Jordan. Graphical models, exponential families, and variational inference. Foundations and Trends in Machine Learning, 1(1–2):1–305, 2008.
  39. [44]Yizhong Wang, Swaroop Mishra, Pegah Alipoormolabashi, Yeganeh Kordi, Amirreza Mirzaei, Atharva Naik, Arjun Ashok, Arut Selvan Dhanasekaran, Anjana Arunkumar, David Stap, et al. Super-naturalinstructions: Generalization via declarative instructions on 1600+ nlp tasks. In Proceedings of the 2022 conference on empirical methods in natural language processing, pages 5085–5109, 2022.
  40. [45]Alex Wang, Yada Pruksachatkun, Nikita Nangia, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel Bowman. Superglue: A stickier benchmark for general-purpose language understanding systems. Advances in neural information processing systems, 32, 2019.
  41. [46]Stephen J Wright and Benjamin Recht. Optimization for data analysis. (No Title), 2022.DOI
  42. [47]Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel Bowman. Glue: A multi-task benchmark and analysis platform for natural language understanding. In Proceedings of the 2018 EMNLP workshop BlackboxNLP: Analyzing and interpreting neural networks for NLP, pages 353–355, 2018.
  43. [48]Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837, 2022.
  44. [49]Mengzhou Xia, Sadhika Malladi, Suchin Gururangan, Sanjeev Arora, and Danqi Chen. LESS: Selecting influential data for targeted instruction tuning. In Forty-first International Conference on Machine Learning, 2024.
  45. [50]Sang Michael Xie, Hieu Pham, Xuanyi Dong, Nan Du, Hanxiao Liu, Yifeng Lu, Percy Liang, Quoc V. Le, Tengyu Ma, and Adams Wei Yu. DoReMi: Optimizing data mixtures speeds up language model pretraining. In Advances in Neural Information Processing Systems, 2023.
  46. [51]An Yang, Baosong Yang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Zhou, Chengpeng Li, Chengyuan Li, Dayiheng Liu, Fei Huang, Guanting Dong, Haoran Wei, Huan Lin, Jialong Tang, Jialin Wang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Ma, Jianxin Yang, Jin Xu, Jingren Zhou, Jinze Bai, Jinzheng He, Junyang Lin, Kai Dang, Keming Lu, Keqin Chen, Kexin Yang, Mei Li, Mingfeng Xue, Na Ni, Pei Zhang, Peng Wang, Ru Peng, Rui Men, Ruize Gao, Runji Lin, Shijie Wang, Shuai Bai, Sinan Tan, Tianhang Zhu, Tianhao Li, Tianyu Liu, Wenbin Ge, Xiaodong Deng, Xiaohuan Zhou, Xingzhang Ren, Xinyu Zhang, Xipin Wei, Xuancheng Ren, Xuejing Liu, Yang Fan, Yang Yao, Yichang Zhang, Yu Wan, Yunfei Chu, Yuqiong Liu, Zeyu Cui, Zhenru Zhang, Zhifang Guo, and Zhihao Fan. Qwen2 technical report, 2024.arxiv.org/abs/2407.10671
  47. [52]Haizhong Zheng, Rui Liu, Fan Lai, and Atul Prakash. Coverage-centric coreset selection for high pruning rates. In 11th International Conference on Learning Representations, ICLR 2023, 2023.
  48. [53]Xiaoyu Zhang, Juan Zhai, Shiqing Ma, Chao Shen, Tianlin Li, Weipeng Jiang, and Yang Liu. Staff: Speculative coreset selection for task-specific fine-tuning. In The thirteenth international conference on learning representations, 2025.

Paper details

vlnc:W_twdhnhxk
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

generative AI tools

View statement
Looking for related papers…
Contents