Pith. sign in

REVIEW 11 cited by

Large-Scale Data Selection for Instruction Tuning

Not yet reviewed by Pith; the record is open.

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 2503.01807 v2 pith:NIPY4YKV submitted 2025-03-03 cs.CL

Large-Scale Data Selection for Instruction Tuning

classification cs.CL
keywords dataselectionsampleslargermethodsmodelspoolsdatasets
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Selecting high-quality training data from a larger pool is a crucial step when instruction-tuning language models, as carefully curated datasets often produce models that outperform those trained on much larger, noisier datasets. Automated data selection approaches for instruction-tuning are typically tested by selecting small datasets (roughly 10k samples) from small pools (100-200k samples). However, popular deployed instruction-tuned models often train on hundreds of thousands to millions of samples, subsampled from even larger data pools. We present a systematic study of how well data selection methods scale to these settings, selecting up to 2.5M samples from pools of up to 5.8M samples and evaluating across 7 diverse tasks. We show that many recently proposed methods fall short of random selection in this setting (while using more compute), and even decline in performance when given access to larger pools of data to select over. However, we find that a variant of representation-based data selection (RDS+), which uses weighted mean pooling of pretrained LM hidden states, consistently outperforms more complex methods across all settings tested -- all whilst being more compute-efficient. Our findings highlight that the scaling properties of proposed automated selection methods should be more closely examined. We release our code, data, and models at https://github.com/hamishivi/automated-instruction-selection.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 11 Pith papers

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

  1. From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning

    cs.LG 2026-05 conditional novelty 7.0

    AutoSelection discovers data recipes from a 90K instruction pool that outperform full-data training and other selectors on reasoning tasks for SFT across multiple models.

  2. Beyond Variance: Prompt-Efficient RLVR via Rare-Event Amplification and Bidirectional Pairing

    cs.LG 2026-02 unverdicted novelty 7.0

    Positive-negative prompt pairing with weighted GRPO improves RLVR sample efficiency, raising AIME 2025 Pass@8 from 16.8 to 22.2 on Qwen2.5-Math-7B while matching large-scale training.

  3. Reinforcement Learning for Reasoning in Large Language Models with One Training Example

    cs.LG 2025-04 accept novelty 7.0

    One training example via RLVR boosts LLM math reasoning from 17.6% to 35.7% average across six benchmarks.

  4. DRIFT: Refining Instruction Data via On-Policy Data Attribution

    cs.LG 2026-06 unverdicted novelty 6.0

    DRIFT applies on-policy influence functions with signed weighting and debiasing to attribute and refine SFT data, raising performance on 7B instruction and reasoning models over prior curation methods.

  5. Interaction-Aware Influence Functions for Group Attribution

    cs.LG 2026-05 conditional novelty 6.0

    Extends influence functions with a second-order pairwise interaction term that improves group attribution accuracy over simple summation on multiple model-dataset pairs and instruction-tuning selection tasks.

  6. HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment

    cs.LG 2026-04 unverdicted novelty 6.0

    HEAL mitigates entropy collapse in few-shot RLVR by selectively adding general-domain data and aligning trajectory-level entropy dynamics, matching full-shot performance with 32 target samples.

  7. GIST: Targeted Data Selection for Instruction Tuning via Coupled Optimization Geometry

    cs.LG 2026-02 unverdicted novelty 6.0

    GIST recovers a task-specific low-dimensional subspace from validation gradients using SVD and scores training examples by their alignment within this coupled subspace for LoRA-based instruction tuning.

  8. A Critical Look at Targeted Instruction Selection: Disentangling What Matters (and What Doesn't)

    cs.LG 2026-02 conditional novelty 6.0

    Only gradient-based (LESS) representations make subset-to-query distance a reliable predictor of instruction-tuning performance; greedy round-robin helps most at small budgets, and random selection is surprisingly com...

  9. Representation-Based Exploration for Language Models: From Test-Time to Post-Training

    cs.LG 2025-10 conditional novelty 6.0

    Representation-based elliptical bonuses improve inference-time and post-training pass@k for LLM reasoning, but the headline AIME result is tainted by validation/test overlap.

  10. Rethinking Data Curation in LLM Training: Online Reweighting Offers Better Generalization than Offline Methods

    cs.LG 2026-04 unverdicted novelty 5.0

    ADAPT is an online reweighting framework for LLM training that outperforms offline data selection and mixing methods in cross-benchmark generalization under equal compute.

  11. Influcoder: Distilling Decoders' Gradient Influence Rankings into an Encoder for Data Attribution

    cs.CL 2026-06 unverdicted novelty 4.0

    Influcoder distills decoders' gradient influence rankings into an encoder for scalable influence-based data attribution.