Skip to content

TTR induction/learning algorithm fully working

Latest

Choose a tag to compare

@github-actions github-actions released this 13 Jul 21:06

[0.2.6] - 2026-07-13

Bug fixes

  • Enhance CI workflow with diagnostics and timeout, pin uv version to avoid regressions

  • Remove ICP del loguru cleanup that deadlocked pytest exit

Documentation

  • Update setup instructions and add quickstart example for parser usage

  • Add contributing guidelines and induction extension notes; update induction pipeline documentation for clarity and structure

Features

  • Implement step-through functionality in parser and UI, including button integration and API updates

  • Implement bounded language derivation functionality and integrate with InteractiveContextParser

  • Integrate loguru for enhanced logging capabilities including silent mode.

  • Add vis method for printing tree in address order and update ParseResult to utilise it

  • Enhance layered language derivation with new output handling and vocabulary functions

  • Add flushing to output handling for immediate file writing

  • Add multi-process parallelisation for layered language derivation.

  • Extend layered language derivation to include fringe output files for each layer

  • Improve status updates in layered language derivation (i/total)

  • Get induction working for three-words input.

  • Enhance TTR functionality with new methods and improved subsumption logic.

  • Add new YAML configurations for various split strategies and enhance CLI for TTR induction pipeline. Introduced holdout, k-fold, pre-split, and train-val-test configurations, and a comprehensive documentation update for usage instructions.

  • Add clean babyDS data.

Bug

  • Remove unused pytest import to fix CI error

  • Implement multiprocessing context for language derivation to avoid fork-related issues

  • Fix CI hang hopefully

Chore

  • Add .gitignore entry for vss directory

  • Drop fixture parity tests; align configs with BabyDS.

Refactor

  • Improve induction algorithm and induction pipeline.

  • Update report generation and documentation for clarity on CLI output

  • Enhance evaluation and hypothesis generation logic (fully matching java now)

  • Update evaluation settings for presplit