Pith. sign in

REVIEW 6 cited by

From Quantity to Quality: Boosting LLM Performance with Self-Guided 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 2308.12032 v5 pith:SVXHQJTP submitted 2023-08-23 cs.CL

From Quantity to Quality: Boosting LLM Performance with Self-Guided Data Selection for Instruction Tuning

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

In the realm of Large Language Models (LLMs), the balance between instruction data quality and quantity is a focal point. Recognizing this, we introduce a self-guided methodology for LLMs to autonomously discern and select cherry samples from open-source datasets, effectively minimizing manual curation and potential cost for instruction tuning an LLM. Our key innovation, the Instruction-Following Difficulty (IFD) metric, emerges as a pivotal metric to identify discrepancies between a model's expected responses and its intrinsic generation capability. Through the application of IFD, cherry samples can be pinpointed, leading to a marked uptick in model training efficiency. Empirical validations on datasets like Alpaca and WizardLM underpin our findings; with a mere $10\%$ of original data input, our strategy showcases improved results. This synthesis of self-guided cherry-picking and the IFD metric signifies a transformative leap in the instruction tuning of LLMs, promising both efficiency and resource-conscious advancements. Codes, data, and models are available: https://github.com/tianyi-lab/Cherry_LLM

discussion (0)

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

Forward citations

Cited by 6 Pith papers

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

  1. Once-For-All: A Train-Once and Select-Anytime Framework for Multimodal Instruction Tuning

    cs.CV 2026-05 unverdicted novelty 7.0

    A selector trained once on LLaVA-665K in CLIP space selects 15% of instructions to reach 98.3% of full-data performance and generalizes to an unseen dataset and different VLMs.

  2. SFGA: A Statistics-First Gating Architecture with Adjudicative Escalation for Trustworthy SFT Data Procurement

    cs.LG 2026-07 conditional novelty 6.0

    A cost-aware gate that routes simple SFT procurement cases to cheap statistics and hard cases to an adjudicated LLM debate reports 0.90 accuracy on a synthetic closed-loop benchmark.

  3. The Ratchet Effect in Silico: How Interaction Drives Cumulative Intelligence in Large Language Models

    cs.LG 2025-07 unverdicted novelty 6.0

    Populations of 1-4B parameter LLMs using peer verification and shared cultural memory achieve 8.8-18.9 point gains on mathematical reasoning tasks and close much of the gap to 70B+ single models.

  4. Muon is Scalable for LLM Training

    cs.LG 2025-02 unverdicted novelty 6.0

    Muon optimizer with weight decay and update scaling achieves ~2x efficiency over AdamW for large LLMs, shown via the Moonlight 3B/16B MoE model trained on 5.7T tokens.

  5. VisNec: Measuring and Leveraging Visual Necessity for Multimodal Instruction Tuning

    cs.CV 2026-03 conditional novelty 5.0

    Selecting instruction-tuning samples by the loss difference between text-only and multimodal prediction (VisNec) lets a model match or exceed full-data performance with only 15% of the data.

  6. Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model

    cs.CV 2025-02 unverdicted novelty 4.0

    Step-Video-T2V describes a 30B-parameter text-to-video model with custom Video-VAE, 3D DiT, flow matching, and Video-DPO that claims state-of-the-art results on a new internal benchmark.