Pith. sign in

REVIEW 26 cited by

LESS: Selecting Influential Data for Targeted Instruction Tuning

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.04333 v3 pith:2BAMEOZM submitted 2024-02-06 cs.CL cs.AIcs.LG

Mengzhou Xia , Sadhika Malladi , Suchin Gururangan , Sanjeev Arora , Danqi Chen This is my paper · ORCID
classification cs.CLcs.AIcs.LG
keywords datainstructionlessmodelseffectivelygradienttuningcapabilities
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Instruction tuning has unlocked powerful capabilities in large language models (LLMs), effectively using combined datasets to develop generalpurpose chatbots. However, real-world applications often require a specialized suite of skills (e.g., reasoning). The challenge lies in identifying the most relevant data from these extensive datasets to effectively develop specific capabilities, a setting we frame as targeted instruction tuning. We propose LESS, an optimizer-aware and practically efficient algorithm to effectively estimate data influences and perform Low-rank gradiEnt Similarity Search for instruction data selection. Crucially, LESS adapts existing influence formulations to work with the Adam optimizer and variable-length instruction data. LESS first constructs a highly reusable and transferable gradient datastore with low-dimensional gradient features and then selects examples based on their similarity to few-shot examples embodying a specific capability. Experiments show that training on a LESS-selected 5% of the data can often outperform training on the full dataset across diverse downstream tasks. Furthermore, the selected data is highly transferable: smaller models can be leveraged to select useful data for larger models and models from different families. Our qualitative analysis shows that our method goes beyond surface form cues to identify data that exemplifies the necessary reasoning skills for the intended downstream application.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 26 Pith papers

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

  1. Bridging Compute- and Data-Optimal Pretraining

    cs.LG 2026-07 conditional novelty 7.0 of 10

    Pretraining loss obeys a single law in which repeated or paraphrased tokens count as η(N, data-per-parameter, expansion-ratio) fresh tokens, with total effective data saturating as derived tokens grow.

  2. Towards Efficient and Effective Alignment of Large Language Models

    cs.CL 2025-06 conditional novelty 7.0 of 10

    A thesis presenting Lion, WebR, LTE, BMC, and FollowBench, five empirical methods that together address LLM alignment data, training, and evaluation.

  3. SafeBuild-Bench: A Temporal-Robust Construction Safety Benchmark with Graph-Enhanced Data Mining

    cs.CV 2026-07 conditional novelty 6.0 of 10

    A temporal-aware construction-safety benchmark built from real inspection archives shows current multimodal models remain far from reliable, with best overall scores near 60.

  4. RSIBench-Data: Benchmarking Data-Centric Research for Recursive Self-Improvement

    cs.SE 2026-07 conditional novelty 6.0 of 10

    A controlled benchmark shows LLM agents can sometimes discover better training-data strategies through feedback, but their improvements are fragile and usually not sustained.

  5. DataShield: Uncovering Risky Fine-Tuning Data Across LLMs Through Consensus Subspace Alignment

    cs.CR 2026-07 conditional novelty 6.0 of 10

    DataShield filters or masks LLM fine-tuning data by measuring alignment with consensus safe/unsafe subspaces from multiple source models, cutting attack success rates after fine-tuning while preserving utility.

  6. GradAlign: Gradient-Aligned Data Selection for LLM Reinforcement Learning

    cs.LG 2026-02 reject novelty 6.0 of 10

    GradAlign selects RL training problems whose policy gradients align with validation-set gradients, claiming more stable and better-performing LLM post-training.

  7. BLISS: A Lightweight Bilevel Influence Scoring Method for Data Selection in Language Model Pretraining

    cs.LG 2025-10 conditional novelty 6.0 of 10

    A bilevel optimization method ranks pretraining data by training a small proxy model on weighted samples, yielding modest downstream-task gains without external pretrained models.

  8. Influence-driven Curriculum Learning for Pre-training on Limited Data

    cs.CL 2025-08 unverdicted novelty 6.0 of 10

    Sorting pre-training examples by training-data influence instead of human-judged difficulty reportedly gives over 10 percentage point benchmark gains over random order in limited-data language model pre-training.

  9. Understanding Data Influence with Differential Approximation

    cs.LG 2025-08 conditional novelty 6.0 of 10

    This paper introduces Diff-In, an influence estimator that accumulates second-order approximations of influence differences across training steps and shows strong accuracy in data cleaning, deletion, and coreset selec...

  10. Approximating Language Model Training Data from Weights

    cs.CL 2025-06 conditional novelty 6.0 of 10

    A gradient-based greedy selection method (SELECT) recovers effective substitute fine-tuning data from two language model checkpoints, approaching the original model's performance on classification and SFT tasks.

  11. EPiC: Towards Lossless Speedup for Reasoning Training through Edge-Preserving CoT Condensation

    cs.LG 2025-06 conditional novelty 6.0 of 10

    Keeping only the first and last parts of chain-of-thought training traces preserves math reasoning accuracy on MATH500 at half the sequence length, but not on harder benchmarks.

  12. SynthRL: Scaling Visual Reasoning with Verifiable Data Synthesis

    cs.LG 2025-06 conditional novelty 6.0 of 10

    SynthRL synthesizes harder, answer-preserving visual math questions from easy seed questions and reports small but mixed out-of-domain RLVR gains for Qwen2.5-VL-7B.

  13. Data Pruning by Information Maximization

    cs.CV 2025-06 conditional novelty 6.0 of 10

    InfoMax selects a training subset by solving a discrete quadratic program that balances sample importance scores against pairwise similarity penalties, and reports state-of-the-art pruning results across three deep le...

  14. Reasoning Like an Economist: Post-Training on Economic Problems Induces Strategic Generalization in LLMs

    cs.AI 2025-05 conditional novelty 6.0 of 10

    Post-training an LLM on economic reasoning problems improves accuracy on economic benchmarks and, without game-specific training, raises its Nash equilibrium frequency and win rates in strategic games.

  15. Daunce: Data Attribution through Uncertainty Estimation

    cs.LG 2025-05 conditional novelty 6.0 of 10

    DAUNCE computes training-data attribution as the covariance of per-example losses across an ensemble of perturbed fine-tuned models, reporting state-of-the-art LDS scores and the first attribution runs on proprietary LLMs.

  16. Merge to Mix: Mixing Datasets via Model Merging

    cs.LG 2025-05 conditional novelty 6.0 of 10

    Merge to Mix shows that the performance of a parameter-averaged model predicts the performance of a model fine-tuned on any dataset mixture, enabling fast and accurate dataset mixture selection.

  17. Less Data, Better Alignment: Data-Centric Multi-Evaluator Agreement for Preference Optimization

    cs.AI 2026-07 conditional novelty 5.0 of 10

    On-policy multi-evaluator consensus filtering retains ~3.5% of candidates and, with KTO on 1.9k labels, improves Mistral-7B preference metrics over stronger data-heavy baselines, with mixed transfer to Llama.

  18. DomainPilot: Domain-Level Loss-Guided Two-Stage Data Mixture Optimization for Efficient Language Model Fine-Tuning

    cs.LG 2026-07 conditional novelty 5.0 of 10

    Reallocating fine-tuning data proportions according to fitted per-domain loss curves improved benchmark scores by 1.8–3.8 points on Qwen3-1.7B, in a single unseeded run; the planned second stage is not yet implemented.

  19. DataPrep-Bench: Benchmarking LLMs as Training Data Preparators

    cs.LG 2026-05 conditional novelty 5.0 of 10

    DataPrep-Bench jointly benchmarks data construction and data-quality evaluation for LLMs across six domains with downstream fine-tuning performance as ground truth.

  20. Active Domain Knowledge Acquisition with 100-Dollar Budget: Enhancing LLMs via Cost-Efficient, Expert-Involved Interaction in Sensitive Domains

    cs.CL 2025-08 unverdicted novelty 5.0 of 10

    A budget-aware framework (PU-ADKA) selects which domain expert an LLM should query under a fixed $100 budget, improving specialized-domain answers at low cost.

  21. ADMIRE-BayesOpt: Accelerated Data MIxture RE-weighting for Language Models with Bayesian Optimization

    stat.ML 2025-08 conditional novelty 5.0 of 10

    Using Bayesian optimization over Gaussian-process surrogates, data mixtures for LLM training can be found much faster than with linear or exponential regression baselines, including across model sizes.

  22. Efficient Data Selection at Scale via Influence Distillation

    cs.CL 2025-05 conditional novelty 5.0 of 10

    Influence Distillation selects LLM fine-tuning data by approximating each sample's gradient influence on a target task via landmarks and JVP embeddings, matching or beating RDS+ accuracy at roughly one third the selec...

  23. SDO: Structure-Aware Data Organization for Efficient LLM Post-Training

    cs.LG 2026-07 conditional novelty 4.0 of 10

    SDO accelerates LLM post-training by forming coherent KNN-based mini-batches and dynamically rebalancing sample exposure across epochs.

  24. Class-Proportional Coreset Selection for Difficulty-Separable Data

    cs.LG 2025-07 conditional novelty 4.0 of 10

    Class-proportional variants of difficulty-based coreset selection outperform class-agnostic methods on class-imbalanced security and medical datasets, particularly at 90-99.9% pruning rates.

  25. Attributing Data for Sharpness-Aware Minimization

    cs.LG 2025-07 reject novelty 4.0 of 10

    SAM-HIF and SAM-GIF are proposed as data attribution scores for SAM-trained models, but SAM-GIF is TracIn with SAM gradients and SAM-HIF's derivation contains a load-bearing error.

  26. Data Diversification Methods In Alignment Enhance Math Performance In LLMs

    cs.AI 2025-07 reject novelty 4.0 of 10

    DTS, which generates diverse solution strategies before writing solutions, improves GSM8K by 7.1 points and MATH by 4.2 points over an untuned base model at 1.03x baseline compute.

Pith tools