Skip to content

Releases: RishanthRajendhran/IdeaLens

idealens 0.1.6

Choose a tag to compare

@RishanthRajendhran RishanthRajendhran released this 06 Oct 03:32

Cites the paper (arXiv:2610.06778): the README has the citation, idealens --help prints the BibTeX, idealens.CITATION holds it, and the PyPI page links the paper. No other changes.

idealens 0.1.5

Choose a tag to compare

@RishanthRajendhran RishanthRajendhran released this 04 Oct 02:16

Links the IdeaLens & ProseLens demo (http://ideadetector.ai/) from the README, docs/USAGE.md, paper/README.md and the PyPI project links. No code changes.

idealens 0.1.4

Choose a tag to compare

@RishanthRajendhran RishanthRajendhran released this 02 Oct 17:13

License change: the code, prompts and results are now released under CC BY-NC-SA 4.0 (versions 0.1.0 to 0.1.3 were Apache 2.0). No code changes.

idealens 0.1.3

Choose a tag to compare

@RishanthRajendhran RishanthRajendhran released this 01 Oct 23:44
  • IdeaLens and ProseLens can now be scored on Tinker without a GPU: pip install "idealens[tinker]", set TINKER_API_KEY, add --backend tinker. Both checkpoints are published on Tinker; scoring is billed per token to your Tinker account (about $0.50 per 1,000 documents for IdeaLens at list price).
  • --dry-run now prices Tinker scoring in dollars.
  • Fix: idealens run --dry-run failed when --backend was left to the model.

idealens 0.1.2

Choose a tag to compare

@RishanthRajendhran RishanthRajendhran released this 01 Oct 23:11

Documentation release: links the IdeaLens Hugging Face collection (all 12 models and 4 datasets) from the README and the package metadata. No code changes.

idealens 0.1.1

Choose a tag to compare

@RishanthRajendhran RishanthRajendhran released this 01 Oct 19:19

Documentation release: the README now explains every way to use the package and the detectors (which detector to pick, end-to-end and step-by-step runs, scoring your own outlines or documents, choosing the extraction LLM including local servers, giving formats or classifying locally, backends, operating points and calibration, large jobs), links every model to its Hugging Face page, and corrects the Qwen weight size (about 16 GB). No code changes.

idealens 0.1.0

Choose a tag to compare

@RishanthRajendhran RishanthRajendhran released this 01 Oct 18:59

First release of the idealens package: classify, extract and score documents with IdeaLens, ProseLens and the baselines. Install with pip install idealens; see the README for models, data and usage.