# HongyuanLuke/frequencylaw

Official repository for textual frequency law

Repository: https://github.com/HongyuanLuke/frequencylaw
Canonical: https://ross.abutalabs.com/products/frequencylaw
Language: Python
License Family: other
Last push: 2026-04-14T12:24:58+00:00

## Health v2 (maintenance only)
Score: 49/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 77, release rhythm 35, longevity 11
- inputs: {"age_days": 160, "days_push": 141, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1435, forks 38 (observed 2026-08-28T04:04:43.512568+00:00)

## What it is
Official code repository for the paper 'Textual Frequency Law on Large Language Models', implementing TFL, TFD, and CTFT methods for studying how textual frequency affects LLM performance. It provides end-to-end scripts for frequency calculation, paired dataset construction, LoRA fine-tuning, and evaluation on GSM8K math reasoning and FLORES-200 translation tasks.

## Use cases
- reproduce textual frequency law experiments on LLMs
- build high/low frequency paired datasets for math reasoning
- fine-tune LLMs with curriculum frequency-based training
- evaluate LLM performance on GSM8K and FLORES-200
- calculate textual frequency statistics for training data
- study how token frequency impacts LLM reasoning and translation

## When to choose
- you want to reproduce or extend the paper's frequency-based LLM training methods
- you need frequency-sorted or paired datasets for reasoning/translation experiments
- you're researching how textual frequency affects model optimization

## When to avoid
- you need a production-ready LLM training framework
- you want a general-purpose fine-tuning tool unrelated to frequency research
- you need a maintained library with a license and long-term support

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-training, nlp, data-generation, benchmarking
- domain: large-language-models, machine-learning
- platform: python
- tags: llm-research, fine-tuning, lora, mathematical-reasoning, machine-translation, academic-paper, dataset-construction, gsm8k, flores-200, natural-language-processing, research

## Member repositories
- HongyuanLuke/frequencylaw (main) score 49

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:43.512568+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:36:50.088410+00:00, confidence not recorded.
  - readme: https://github.com/HongyuanLuke/frequencylaw (fetched 2026-08-28T04:04:43.512568+00:00, sha 630ad9afa898)
- Data as of 2026-08-30T08:39:29.467469+00:00.