MIG: Automatic Data Selection for Instruction Tuning by Maximizing Information Gain in Semantic Space

Haochen Ye; Kai Chen; Kai Hu; Yicheng Chen; Yining Li; Zerun Ma

REVIEW 4 major objections 6 minor 1 cited by

MIG: Automatic Data Selection for Instruction Tuning by Maximizing Information Gain in Semantic Space

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims that greedily maximizing information gain over a semantic label graph selects instruction-tuning data so well that 5% of the pool matches full-data fine-tuning.

desk verdict Novel label-graph selection idea, but the published algorithm is not executable and the theory overclaims; worth peer review after a major fix. read the letter →

arxiv 2504.13835 v1 pith:IND5BG7B submitted 2025-04-18 cs.CL cs.AI

Yicheng Chen , Yining Li , Kai Hu , Zerun Ma , Haochen Ye , Kai Chen This is my paper · ORCID

classification cs.CLcs.AI

keywords instructiontuningdataselectionsubmodularmaximizationinformationgainlabelgraphsemanticdiversityqualityLLMfine-tuning

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

The paper proposes a single quantitative measure for judging an instruction-tuning dataset: total information distributed over a semantic label graph, where each sample contributes its quality score to the labels it carries, and a concave function makes the measure submodular so greedy selection is near-optimal. The resulting algorithm, MIG, iteratively adds the sample with the largest approximate information gain, and the paper reports that this rule selects subsets that consistently outperform existing quality-and-diversity samplers across three data pools and three base models. Most strikingly, fine-tuning on 5% of the Tulu3 pool selected by MIG yields performance comparable to the official SFT model trained on the full 939K-record pool, with gains of +5.73% on AlpacaEval and +6.89% on Wildbench. If correct, this turns instruction-data curation from a heuristic art into a form of submodular optimization with a cheap, 100x-faster sampler.

What carries the argument

The load-bearing object is the labeled semantic graph $G_L=(L,E_L)$ with weighted edges $w_{pq}$ from textual similarity (Eq. 3), the propagation matrix $A$ from Eq. (6)-(7), and the concave information score $\phi$. The dataset measure $E(D)=\Phi(A\sum_{i\in D} s_i v_i)$ is submodular (Appx. C), which licenses the greedy sampler of Alg. 1; the key computational step is the gradient approximation $G_k=A\Phi'(A\sum_{i\in D_k} e_i)$ of the exact marginal gain, which avoids recomputing the whole objective and reduces selection cost to $O(NK)$ over labels rather than pairwise distances over the pool.

What would settle it

Run MIG on a ~10K-record pool once with the exact marginal gain of Eq. (9) and once with the gradient approximation of Eq. (10); if the selected subsets differ appreciably in composition or downstream benchmark scores, then the implemented algorithm is not the submodular greedy method whose (1-1/e) guarantee is claimed.

Watch

Extended reading notes

Core claim

The central claim is that the information content of an instruction-tuning dataset can be modeled as the aggregate of per-label information over a graph whose nodes are semantic labels and whose edges encode label similarity, with each data point $d_i$ depositing its quality score $s_i$ into its associated labels. Dataset information is $E(D)=\Phi(A\sum_{i\in D} s_i v_i)$ where $v_i$ is the binary label vector, $A$ is a label-propagation matrix, and $\Phi$ applies an increasing concave function $\phi(x)=x^{0.8}$ elementwise, so that adding more data to an already information-rich label yields diminishing returns. This makes $E$ submodular, and MIG is the greedy algorithm that at each step selects the candidate with the largest marginal gain, computed via the gradient $G_k=A\Phi'(A\sum_{i\in D_k} e_i)$. The paper's empirical claim is that the resulting subsets dominate prior selection methods on knowledge-based and human-preference benchmarks, and that a 5% Tulu3 sample matches or exceeds full-data SFT, demonstrating that a global semantic measure with a greedy optimizer can replace heuristic diversity filters and pairwise embedding distances.

Load-bearing premise

The implemented greedy sampler assumes that the first-order gradient in Eq. (10) ranks candidate data points the same way the exact submodular marginal gain in Eq. (9) would, and the paper provides no error bound for that approximation.

Editorial extensions

If this is right

A 5% Tulu3 subset sampled by MIG (about 47K of 939K records) yields a Llama3.1-8B model whose nine-benchmark average matches the official full-data SFT, with human-preference scores up by +4.59%; dataset-size reduction of this magnitude changes the economics of post-training.
MIG's gains hold across Llama3.1-8B, Mistral-7B-v0.3, and Qwen2.5-7B and across Tulu3, Openhermes2.5, and Xsota, so the selection rule transfers without retuning per model family.
Sampling 50K records from Tulu3 takes about 0.45 GPU-hours with MIG versus 81-86 GPU-hours for DEITA and QDIT; removing the pairwise-distance bottleneck makes submodular selection practical at million-record scale.
The grid-search sweeps show unimodal performance ridges in label-set size and edge density, indicating that for each pool there is an optimal graph granularity, and that MIG's information-propagation weight $\alpha=1.0$ outperforms no propagation by +2.76 average points.

Reading between the lines

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

Because the implementation replaces the exact submodular gain with a first-order gradient, the (1-1/e) guarantee formally applies to $E$ itself, not to the algorithm actually run; a direct comparison of exact-gain and gradient-gain selections on a small pool would tell how much of MIG's empirical edge comes from submodularity and how much from the DEITA quality scores that seed it.
The paper treats the label tagger and embedding model as fixed; since its own parameter sweeps show sensitivity to label granularity, an obvious extension is to co-train or adapt the label graph per data pool rather than grid-searching thresholds.
MIG's largest gains appear on human-preference benchmarks (Wildbench, AlpacaEval) rather than knowledge benchmarks, suggesting the same information-gain objective could serve as a data ordering or filtering step before RLHF, not only for SFT.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 central idea here is genuinely new: model the instruction space as a label graph, propagate information along edges, and greedily maximize a concave information gain. That mechanism is a real departure from independent-label or embedding-distance methods, and the submodularity proof in Appx C is correct for the exact objective. The efficiency argument is plausible too: a graph over labels avoids the pairwise embedding costs of QDIT and DEITA. If the empirical claims hold, the 5% Tulu3 result is a useful data point.

But the paper has a load-bearing problem. The published Algorithm 1 is not executable as written. Eq. (10) defines G = A Φ'(A Σ e_i), but the correct first-order marginal gain for adding a data point is (A^T Φ'(z))·e_d. The propagation matrix A in Eq. (6) is not symmetric in general—the denominator depends on the row index—so the algorithm as written does not compute the gradient it claims. Worse, with Φ(x)=x^0.8, Φ'(0)=∞, so at the first iteration (D_S empty) the score vector G is all infinities and the argmax in Eq. (11) is undefined. No epsilon or stabilization is mentioned. The selection rule that produced Tables 1–2 is therefore not specified in the paper. The (1-1/e) guarantee from Appx C applies to the exact greedy marginal gain, not to the gradient approximation, and no error bound is given for that approximation. These aren't cosmetic typos; they break the link between the theory, the algorithm, and the reported results.

The empirical work is broad but not airtight. Hyperparameters (data size, epochs, graph parameters) are tuned on the same evaluation benchmarks used for the headline numbers—Table 3 is a grid search for the exact setup that produces the main results. All results are single runs, no variance estimates, and some margins are small (e.g., +0.41 on Openhermes2.5). The limitation section honestly admits the grid-search dependence, but that doesn't remove the circularity.

Bottom line: the formulation could become a real contribution, but this manuscript is not reproducible and the theoretical claims outrun what the algorithm actually implements. I would send it to peer review—the idea deserves referee attention—but with an expectation of major revision: correct the gradient (transpose, finite derivatives, an epsilon), release the code, add multi-seed runs, and move hyperparameter selection to a held-out split. As it stands, I wouldn't cite it, and I'd advise a student to read it for the label-graph idea but not to trust the reported gains.

Referee Report

4 major / 6 minor

Summary. The paper proposes MIG, an automatic data-selection method for instruction tuning. The method annotates each data point with a set of labels and a quality score, builds a weighted label graph, and defines a set function E(D) = Φ(A Σ_{i∈D} s_i v_i) that aggregates propagated label information through a concave, monotonically increasing elementwise function. The paper proves that E is submodular when the elementwise function is concave and non-decreasing, and then presents a greedy sampling algorithm that approximates the marginal information gain by a first-order gradient expression. Experiments on Tulu3, Openhermes2.5, and Xsota pools with Llama3.1-8B, Mistral-7B-v0.3, and Qwen2.5-7B report consistent improvements over baselines, including the headline result that 5% of Tulu3 data selected by MIG matches or exceeds the official full-data SFT model on human-preference benchmarks.

Significance. If correct, MIG would provide a principled and computationally efficient method for instruction-data selection, combining a submodular objective with a graph-based semantic model and a roughly 100-fold speedup over embedding-based alternatives. The submodularity proof in Appendix C is mathematically sound for the stated conditions, and the experimental scope is broad, spanning three data pools, three base models, and nine evaluation benchmarks. However, the paper's central methodological claims are compromised by serious gaps between the theory, the written algorithm, and the actual implementation: the gradient update in Eq. (10) is not the first-order marginal gain unless the propagation matrix is symmetric, the chosen φ(x)=x^{0.8} has an undefined derivative at the initial empty set, and no approximation error analysis connects the implemented gradient step to the (1-1/e) greedy guarantee. The reported empirical gains are also partly fitted because the key hyperparameters are tuned on the same evaluation benchmarks used for the main comparison. The underlying idea is promising, but the manuscript as written does not specify a reproducible selection rule.

major comments (4)

[§3.3, Eq. (10), Algorithm 1, Eq. (6)] The score G·e_d used in the algorithm is not the first-order approximation of the exact marginal gain in Eq. (9). For z = A Σ_{i∈D_S} e_i, the true first-order gain of adding d is e_d^T A^T Φ'(z), whereas Algorithm 1 computes Φ'(z)^T A^T e_d by taking G = AΦ'(z) and then G·e_d. Since the propagation matrix A defined in Eq. (6) is not symmetric (the denominator is indexed by the source label p), these two quantities differ. Consequently, the greedy algorithm whose submodularity guarantee is proved in Appendix C is not the algorithm that is executed. The authors should either correct the transpose, restrict A to be symmetric, or prove that the two expressions produce the same ranking of candidates.
[§4.3 and Appendix A.2] The chosen information score function φ(x) = x^{0.8} has φ'(0) = ∞. At the first iteration of Algorithm 1, D_S is empty, so z = 0 and the vector G in Eq. (10) is a vector of infinities; the argmax in Eq. (11) is therefore undefined, and any data point that touches any label receives an infinite score. The manuscript does not state an epsilon offset, an initialization step, or any stabilization of the derivative. The selection rule that actually produced Tables 1 and 2 is consequently not specified, making the central empirical claim non-reproducible from the paper as written.
[Appendix C.2] The (1-1/e) approximation guarantee of Nemhauser et al. applies to a greedy algorithm that evaluates the exact marginal gain of a submodular function. Algorithm 1 replaces that exact gain with the gradient approximation of Eq. (10), and no bound is given on the error between the approximate score and the true marginal gain. The paper therefore invokes a theoretical guarantee that does not formally cover the implemented method. Please either run the exact greedy algorithm (which is feasible with the graph-based objective), derive an approximation guarantee for the gradient-based selection, or present the gradient step explicitly as a heuristic without claiming the submodularity guarantee.
[§4.3, Table 3, Fig. 4-6, Limitation] The hyperparameters that define MIG (sample budget, training epochs, φ exponent, propagation weight α, label set size, and edge threshold T) are selected by grid search on the same Tulu3 evaluation benchmarks that are later used to report the main results in Table 1. The Limitation paragraph acknowledges that the parameters depend on grid search, but the headline claim that MIG 'consistently outperforms' on Tulu3 is partly a fitted result. To support the generalization claim, the authors should either fix the parameters a priori across data pools, tune on a held-out validation set, or demonstrate that the reported improvements persist without per-pool tuning on the test benchmarks.

minor comments (6)

[§3.2, Eq. (3)] The symbol σ is used in Eq. (3) without definition; from context it appears to be an indicator function, but this should be stated explicitly.
[§3.2] The phrase 'upper-convex function' is inconsistent with the stated 'marginally diminishing information gain' and with the concavity assumed in Appendix C; the terminology should be corrected to 'concave' or 'with diminishing returns'.
[§3.2, Eq. (6)] In the definition of the propagation weight a_{pq}, the quantity w_p is said to 'equal 1', which does not match its role as a normalizer in the denominator; the definition of w_p should be clarified.
[Algorithm 1, line 4] The notation E_k in Algorithm 1 is not defined and is inconsistent with e_i used in Eq. (10); please use consistent notation for the propagated information vectors.
[§4.3, Eq. (12)] The exponential candidate Φ(x) = 1 - e^{-αx} is introduced but no experimental results are reported for it; if it was evaluated, the results should be included, and if not, the equation should be removed or clearly marked as unused.
[Table 3] The grid search table reports only the Avg score; please also report Avg_obj and Avg_sub, and ideally run multiple random seeds to provide a sense of variance, since the differences between some cells are small.

Circularity Check

1 steps flagged · score 4.0 of 10

MIG's information objective is not defined from the target benchmarks, so the core derivation is not circular; however, the headline Tulu3 results are partly fitted, because the data budget, epochs, score function, propagation weight, and label graph are all selected by maximizing the same benchmark scores that are then reported as evidence.

fitted input called prediction [Section 4.3 'Grid Search', Table 3; also Figures 4-6 and Appendix A.2 implementation choices]
"All methods select 50K samples based on the grid search (Sec 4.3). ... Results in Table 3 indicate 50K samples with three training epochs as optimal for Tulu3, consistently maximizing performance for MIG and random selection. ... with Φ(x) = x0.8 achieving the best results on human-preference and knowledge-based benchmarks."

The reported MIG Tulu3 result (Avg 55.32 in Table 1) is produced with the exact configuration chosen by the same Tulu3 benchmark averages: 50K samples, 3 epochs, Φ(x)=x^0.8, α=1.0, label-graph size 4531, threshold 0.9, and DEITA quality scores. The abstract's claim that '5% Tulu3 data sampled by MIG achieves comparable performance' is therefore a selected optimum rather than an independent prediction: the fitted hyperparameters are evaluated on the same targets used to select them. This does not make E(D) itself circular, because the information measure is not constructed from AlpacaEval/WildBench outcomes, but it does mean the headline quantitative superiority is partly forced by the search, not derived from the method.

full rationale

The core construction E(D)=Φ(AΣ_{i∈D} s_i v_i) is an independent proxy: quality scores and a label graph are inputs, and the selection objective is not written in terms of the evaluation benchmarks. The submodularity proof in Appx. C is a standard concave-composition argument and does not assume the conclusion. There is no load-bearing self-citation chain: the use of CompassJudger for evaluation is self-referential but not a derivation input. The only circularity-adjacent step is the grid search on the evaluation benchmarks, which makes the headline numbers fitted rather than predicted. The apparent A^T/Φ'(0) issues in Eq. (10) and Algorithm 1 are correctness and reproducibility defects, not circularity, since they break the chain instead of closing it.

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

The central method rests on five free parameters, all tuned on the evaluation benchmarks used for the headline results, and five assumptions, two of which are ad hoc to this paper. The label graph is a construct over existing labels and similarities; no new physical entities are introduced. The most fragile assumption is the gradient approximation of the exact submodular gain, which breaks the formal greedy guarantee.

free parameters (5)

Information score function exponent α (φ(x)=x^α) = 0.8
Chosen by grid search over candidate concave functions on the Tulu3 benchmarks (Fig 4b,c); affects the quality-diversity tradeoff.
Information propagation weight α (Eq. 6) = 1.0
Selected from a sweep (Fig 6c) as the value maximizing Avg on Tulu3.
Edge similarity threshold T (Eq. 3) = 0.9
Selected from threshold sweep (Fig 6b); controls graph density.
Label set size (number of graph nodes) = 4531 (Tulu3), 3059 (Xsota), 5166 (Openhermes2.5)
Chosen per data pool based on experiments (Fig 6a); authors state size is positively correlated with pool size.
Sampling budget N and training epochs = 50K samples, 3 epochs
Selected via grid search in Table 3 to maximize Avg on Tulu3 for both random and MIG; this is a training setting, but it is tuned on the evaluation benchmarks.

assumptions (5)

standard math The dataset information E(D) is submodular when φ is concave non-decreasing and A is a non-negative matrix.
Proven in Appx C using concavity of φ; standard result for concave functions of modular functions.
domain assumption Textual similarity between label names, computed by an embedding model, captures semantic relationships relevant to instruction diversity.
Label graph edges use E5-Mistral embedding cosine similarity (Sec 3.2, Eq 3); no validation that this similarity reflects the semantic structure of instructions.
ad hoc to paper Information propagation rule in Eq. (6) correctly models semantic overlap and annotation bias among labels.
The propagation matrix A is a heuristic design with no derivation or external justification; its row normalization properties are not analyzed.
ad hoc to paper The first-order Taylor approximation G_k e_d of the marginal information gain ranks candidates as accurately as the true gain.
Alg. 1 and Eq. (10)-(11) replace the exact submodular marginal gain with a gradient; no error bound is provided, so the approximation's validity is assumed.
domain assumption A data point's information contribution is proportional to its DEITA quality score and is distributed uniformly over its labels.
Eq. (5) sets e_i = s_i v_i; this assumes quality scales linearly and adds equally across all labels of an example.

how reviews work

0 comments

Cite this review

Pith. "Pith review of MIG: Automatic Data Selection for Instruction Tuning by Maximizing Information Gain in Semantic Space." pith.science (2026). https://pith.science/paper/IND5BG7B

@misc{pith2026250413835,
  author       = {Pith},
  title        = {Pith review of: MIG: Automatic Data Selection for Instruction Tuning by Maximizing Information Gain in Semantic Space},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IND5BG7B}},
  note         = {Machine review of arXiv:2504.13835}
}

read the original abstract

Data quality and diversity are key to the construction of effective instruction-tuning datasets. % With the increasing availability of open-source instruction-tuning datasets, it is advantageous to automatically select high-quality and diverse subsets from a vast amount of data. % Existing methods typically prioritize instance quality and use heuristic rules to maintain diversity. % However, this absence of a comprehensive view of the entire collection often leads to suboptimal results. % Moreover, heuristic rules generally focus on distance or clustering within the embedding space, which fails to accurately capture the intent of complex instructions in the semantic space. % To bridge this gap, we propose a unified method for quantifying the information content of datasets. This method models the semantic space by constructing a label graph and quantifies diversity based on the distribution of information within the graph. % Based on such a measurement, we further introduce an efficient sampling method that selects data samples iteratively to \textbf{M}aximize the \textbf{I}nformation \textbf{G}ain (MIG) in semantic space. % Experiments on various datasets and base models demonstrate that MIG consistently outperforms state-of-the-art methods. % Notably, the model fine-tuned with 5\% Tulu3 data sampled by MIG achieves comparable performance to the official SFT model trained on the full dataset, with improvements of +5.73\% on AlpacaEval and +6.89\% on Wildbench.

Figures

Figures reproduced from arXiv: 2504.13835 by the authors.

**Figure 1.** Comparison with different data selection methods (Lu et al., 2024; Liu et al., 2024b) on the Tulu3 (Lambert et al., 2024) pool using Llama3.1-8B (Touvron et al., 2023), evaluated on (black) knowledge-based benchmarks and (red) human-preference benchmarks. See details in Sec. 4.2. 2020; Taori et al., 2023; Touvron et al., 2023). Instruction tuning utilizes instruction-response pairs to guide base models toward more… view at source ↗

**Figure 2.** Illustration of (a) Data Selection Pipeline and (b) MIG Sampler. Given the raw data pool, our pipeline [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Data scaling experiments on Tulu3 using Llama3.1- 8B. The score reported here is the Avg score. and quality. MIG consistently outperforms all baselines, achieving Avg improvements of +0.41% and +0.99% over previous best methods, further demonstrating its generalizability. Notably, on Xsota, all baselines exhibit performance degradation on knowledge-based evaluations, consistent with the findings in (Xia et al., 20… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: (a) Derivative of Information Score Functions. (b) Avgobj on Different Information Score Functions. (c) Avgsub on Different Quality Scores [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Quantitative results on different quality metrics. DEITA scores achieve the best performance on both humanpreference and knowledge-based evaluations. shown in [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: Analysis of Parameters in the Label Graph. The reported score is the average of Avgsub and Avgobj. Please refer to [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

GemMaroc: Unlocking Darija Proficiency in LLMs with Minimal Data
cs.CL 2025-05 conditional novelty 5.0 of 10

Fine-tuning Gemma 3-4B and 27B on about 50,000 mixed Darija and English instructions produces a 27B model that matches Atlas-Chat on DarijaMMLU and exceeds it on DarijaHellaSwag, using 48 GPU-hours.

Reference graph

Works this paper leans on

51 extracted references · 21 canonical work pages · cited by 1 Pith paper

[1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...
[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]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gr...

work page 2020
[4]

Alexander Bukharin, Shiyang Li, Zhengyang Wang, Jingfeng Yang, Bing Yin, Xian Li, Chao Zhang, Tuo Zhao, and Haoming Jiang. 2024. Data diversity matters for robust instruction tuning. In EMNLP

work page 2024
[5]

Maosong Cao, Alexander Lam, Haodong Duan, Hongwei Liu, Songyang Zhang, and Kai Chen. 2024 a . Compassjudger-1: All-in-one judge model helps model evaluation and evolution. arXiv preprint arXiv:2410.16256

arXiv 2024
[6]

Yihan Cao, Yanbin Kang, Chi Wang, and Lichao Sun. 2024 b . Instruction mining: Instruction data selection for tuning large language models. In COLM

work page 2024
[7]

Lichang Chen, Shiyang Li, Jun Yan, Hai Wang, Kalpa Gunaratna, Vikas Yadav, Zheng Tang, Vijay Srinivasan, Tianyi Zhou, Heng Huang, and Hongxia Jin. 2024. Alpagasus: Training a better alpaca with fewer data. In ICLR

2024
[8]

Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhail Pavlov, Alethea Power, Lukasz Kaiser, Mohammad Bavarian...

arXiv 2021

Show all 51 references

[9]

Gonzalez, Ion Stoica, and Eric P

Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E. Gonzalez, Ion Stoica, and Eric P. Xing. 2023. https://lmsys.org/blog/2023-03-30-vicuna/ Vicuna: An open-source chatbot impressing gpt-4 with 90\

2023
[10]

Peter Clark, Isaac Cowhey, Oren Etzioni, Tushar Khot, Ashish Sabharwal, Carissa Schoenick, and Oyvind Tafjord. 2018. Think you have solved question answering? try arc, the ai2 reasoning challenge. arXiv preprint arXiv:1803.05457

2018 arXiv
[11]

Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168

2021 arXiv
[12]

OpenCompass Contributors. 2023. Opencompass: A universal evaluation platform for foundation models. https://github.com/open-compass/opencompass

2023
[13]

G \'e rard Cornu \'e jols, George Nemhauser, and Laurence Wolsey. 1983. The uncapicitated facility location problem. Technical report, Cornell University Operations Research and Industrial Engineering

1983
[14]

Ning Ding, Yulin Chen, Bokai Xu, Yujia Qin, Shengding Hu, Zhiyuan Liu, Maosong Sun, and Bowen Zhou. 2023. Enhancing chat language models by scaling high-quality instructional conversations. In EMNLP

2023
[15]

Guanting Dong, Hongyi Yuan, Keming Lu, Chengpeng Li, Mingfeng Xue, Dayiheng Liu, Wei Wang, Zheng Yuan, Chang Zhou, and Jingren Zhou. 2024. How abilities in large language models are affected by supervised fine-tuning data composition. In ACL

2024
[16]

Yann Dubois, Bal \'a zs Galambosi, Percy Liang, and Tatsunori B Hashimoto. 2024. Length-controlled alpacaeval: A simple way to debias automatic evaluators. arXiv preprint arXiv:2404.04475

2024 arXiv
[17]

Yuan Ge, Yilun Liu, Chi Hu, Weibin Meng, Shimin Tao, Xiaofeng Zhao, Mahong Xia, Zhang Li, Boxing Chen, Hao Yang, Bei Li, Tong Xiao, and JingBo Zhu. 2024. Clustering and ranking: Diversity-preserved instruction selection through expert-aligned quality estimation. In EMNLP

2024
[18]

Michael Hahsler, Matthew Piekenbrock, and Derek Doran. 2019. dbscan: Fast density-based clustering with r. Journal of Statistical Software

2019
[19]

Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring massive multitask language understanding. In ICLR

2021
[20]

Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas ...

2023 arXiv
[21]

Miranda, Alisa Liu, Nouha Dziri, Shane Lyu, Yuling Gu, Saumya Malik, Victoria Graf, Jena D

Nathan Lambert, Jacob Morrison, Valentina Pyatkin, Shengyi Huang, Hamish Ivison, Faeze Brahman, Lester James V. Miranda, Alisa Liu, Nouha Dziri, Shane Lyu, Yuling Gu, Saumya Malik, Victoria Graf, Jena D. Hwang, Jiangjiang Yang, Ronan Le Bras, Oyvind Tafjord, Chris Wilhelm, Luc...

2024 arXiv
[22]

Guohao Li, Hasan Abed Al Kader Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. 2023 a . CAMEL : Communicative agents for ''mind'' exploration of large language model society. In NIPS

2023
[23]

Ming Li, Yong Zhang, Shwai He, Zhitao Li, Hongyu Zhao, Jianzong Wang, Ning Cheng, and Tianyi Zhou. 2024 a . Superfiltering: Weak-to-strong data filtering for fast instruction-tuning. In ACL

2024
[24]

Ming Li, Yong Zhang, Zhitao Li, Jiuhai Chen, Lichang Chen, Ning Cheng, Jianzong Wang, Tianyi Zhou, and Jing Xiao. 2024 b . From quantity to quality: Boosting LLM performance with self-guided data selection for instruction tuning. In NAACL

2024
[25]

Yunshui Li, Binyuan Hui, Xiaobo Xia, Jiaxi Yang, Min Yang, Lei Zhang, Shuzheng Si, Junhao Liu, Tongliang Liu, Fei Huang, et al. 2023 b . One shot learning as instruction data prospector for large language models. arXiv preprint arXiv:2312.10302

2023 arXiv
[26]

Bill Yuchen Lin, Yuntian Deng, Khyathi Chandu, Faeze Brahman, Abhilasha Ravichander, Valentina Pyatkin, Nouha Dziri, Ronan Le Bras, and Yejin Choi. 2024. Wildbench: Benchmarking llms with challenging tasks from real users in the wild. arXiv preprint arXiv:2406.04770

2024 arXiv
[27]

Wong, Dongfang Li, Ziyi Wang, Baotian Hu, and Min Zhang

Liangxin Liu, Xuebo Liu, Derek F. Wong, Dongfang Li, Ziyi Wang, Baotian Hu, and Min Zhang. 2024 a . Select IT : Selective instruction tuning for LLM s via uncertainty-aware self-reflection. In NIPS

2024
[28]

Wei Liu, Weihao Zeng, Keqing He, Yong Jiang, and Junxian He. 2024 b . What makes good data for alignment? a comprehensive study of automatic data selection in instruction tuning. In ICLR

2024
[29]

Keming Lu, Hongyi Yuan, Zheng Yuan, Runji Lin, Junyang Lin, Chuanqi Tan, Chang Zhou, and Jingren Zhou. 2024. \#instag: Instruction tagging for analyzing supervised fine-tuning of large language models. In ICLR

2024
[30]

Michel Minoux. 2005. Accelerated greedy algorithms for maximizing submodular set functions. In Optimization Techniques: Proceedings of the 8th IFIP Conference on Optimization Techniques W \"u rzburg, September 5--9, 1977

2005
[31]

George L Nemhauser, Laurence A Wolsey, and Marshall L Fisher. 1978. An analysis of approximations for maximizing submodular set functions—i. Mathematical programming

1978
[32]

Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. In EMNLP

2019
[33]

Mirac Suzgun, Nathan Scales, Nathanael Sch \"a rli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc V Le, Ed H Chi, Denny Zhou, , and Jason Wei. 2022. Challenging big-bench tasks and whether chain-of-thought can solve them. arXiv preprint arXiv:2210.09261

2022 arXiv
[34]

Hashimoto

Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. 2023. Stanford alpaca: An instruction-following llama model. https://github.com/tatsu-lab/stanford_alpaca

2023
[35]

Teknium. 2023. https://huggingface.co/datasets/teknium/OpenHermes-2.5 Openhermes 2.5: An open dataset of synthetic data for generalist llm assistants

2023
[36]

Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation la...

2023 arXiv
[37]

Guan Wang, Sijie Cheng, Xianyuan Zhan, Xiangang Li, Sen Song, and Yang Liu. 2023 a . Openchat: Advancing open-source language models with mixed-quality data. arXiv preprint arXiv:2309.11235

2023 arXiv
[38]

Liang Wang, Nan Yang, Xiaolong Huang, Linjun Yang, Rangan Majumder, and Furu Wei. 2023 b . Improving text embeddings with large language models. arXiv preprint arXiv:2401.00368

2023 arXiv
[39]

Peiqi Wang, Yikang Shen, Zhen Guo, Matthew Stallone, Yoon Kim, Polina Golland, and Rameswar Panda. 2024. Diversity measurement and subset selection for instruction tuning datasets. arXiv preprint arXiv:2402.02318

2024 arXiv
[40]

Shengguang Wu, Keming Lu, Benfeng Xu, Junyang Lin, Qi Su, and Chang Zhou. 2023. Self-evolved diverse data sampling for efficient instruction tuning. arXiv preprint arXiv:2311.08182

2023 arXiv
[41]

Mengzhou Xia, Sadhika Malladi, Suchin Gururangan, Sanjeev Arora, and Danqi Chen. 2024 a . LESS : Selecting influential data for targeted instruction tuning. In ICML

2024
[42]

Tingyu Xia, Bowen Yu, Kai Dang, An Yang, Yuan Wu, Yuan Tian, Yi Chang, and Junyang Lin. 2024 b . Rethinking data selection at scale: Random selection is almost all you need. arXiv preprint arXiv:2410.09335

2024 arXiv
[43]

An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, Kai Dang, Keming Lu, Keqin Bao, Kexin Yang, Le Yu, Me...

2024 arXiv
[44]

Mingjia Yin , Chuhan Wu , Yufei Wang , Hao Wang , Wei Guo , Yasheng Wang , Yong Liu , Ruiming Tang , Defu Lian , and Enhong Chen . 2024. Entropy Law: The Story Behind Data Compression and LLM Performance . arXiv preprint arXiv:2407.06645

2024 arXiv
[45]

Longhui Yu, Weisen Jiang, Han Shi, Jincheng YU, Zhengying Liu, Yu Zhang, James Kwok, Zhenguo Li, Adrian Weller, and Weiyang Liu. 2024 a . Metamath: Bootstrap your own mathematical questions for large language models. In ICLR

2024
[46]

Simon Yu, Liangyu Chen, Sara Ahmadian, and Marzieh Fadaee. 2024 b . Diversify and conquer: Diversity-centric data selection with iterative refinement. arXiv preprint arXiv:2409.11378

2024 arXiv
[47]

Zhang, and Yongbin Li

Yingxiu Zhao, Bowen Yu, Binyuan Hui, Haiyang Yu, Minghao Li, Fei Huang, Nevin L. Zhang, and Yongbin Li. 2024. Tree-instruct: A preliminary study of the intrinsic relationship between complexity and alignment. In COLING

2024
[48]

Gonzalez, and Ion Stoica

Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. Judging LLM -as-a-judge with MT -bench and chatbot arena. In NIPS

2023
[49]

Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, and Zheyan Luo. 2024. L lama F actory: Unified efficient fine-tuning of 100+ language models. In ACL

2024
[50]

Chunting Zhou, Pengfei Liu, Puxin Xu, Srinivasan Iyer, Jiao Sun, Yuning Mao, Xuezhe Ma, Avia Efrat, Ping Yu, LILI YU, Susan Zhang, Gargi Ghosh, Mike Lewis, Luke Zettlemoyer, and Omer Levy. 2023 a . Lima: Less is more for alignment. In NIPS

2023
[51]

Jeffrey Zhou, Tianjian Lu, Swaroop Mishra, Siddhartha Brahma, Sujoy Basu, Yi Luan, Denny Zhou, and Le Hou. 2023 b . Instruction-following evaluation for large language models. arXiv preprint arXiv:2311.07911

2023 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

[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]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gr...

work page 2020

[4] [4]

Alexander Bukharin, Shiyang Li, Zhengyang Wang, Jingfeng Yang, Bing Yin, Xian Li, Chao Zhang, Tuo Zhao, and Haoming Jiang. 2024. Data diversity matters for robust instruction tuning. In EMNLP

work page 2024

[5] [5]

Maosong Cao, Alexander Lam, Haodong Duan, Hongwei Liu, Songyang Zhang, and Kai Chen. 2024 a . Compassjudger-1: All-in-one judge model helps model evaluation and evolution. arXiv preprint arXiv:2410.16256

arXiv 2024

[6] [6]

Yihan Cao, Yanbin Kang, Chi Wang, and Lichao Sun. 2024 b . Instruction mining: Instruction data selection for tuning large language models. In COLM

work page 2024

[7] [7]

Lichang Chen, Shiyang Li, Jun Yan, Hai Wang, Kalpa Gunaratna, Vikas Yadav, Zheng Tang, Vijay Srinivasan, Tianyi Zhou, Heng Huang, and Hongxia Jin. 2024. Alpagasus: Training a better alpaca with fewer data. In ICLR

2024

[8] [8]

Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, Nick Ryder, Mikhail Pavlov, Alethea Power, Lukasz Kaiser, Mohammad Bavarian...

arXiv 2021