You can deploy this model and start earning money today!

sida/
icalens-pythia-70m-pile10k

Name: icalens-pythia-70m-pile10k
Availability: InStock

icalens

interpretability

independent-component-analysis

activations

2606.11722

NeelNanda/pile-10k

Views

No views yet

ICA Lens for EleutherAI/pythia-70m

This repository contains a fitted ICA Lens for analyzing internal activations of EleutherAI/pythia-70m. It provides layer-wise ICA transformations for mapping residual-stream activations to independent-component scores and energy shares.

Artifact summary

Field	Value
Analyzed model	EleutherAI/pythia-70m
Analyzed model revision	`a39f36b100fe8a5377810d56c3f4789b9c53ac42`
Model kind	`base`
Activation site	`resid_post`
Layer indexing	`transformer_blocks_zero_based`
Available layers	0, 1, 2, 3, 4, 5
Hidden size	512
Input row normalization	`none`
ICALens package version	`0.3.6`
Fitting dataset	NeelNanda/pile-10k
Dataset revision	`127bfedcd5047750df5ccf3a12979a47bfa0bafa`
Dataset split	`train`
Fitting token scope	`all`
Candidate tokens	5317136
Fitting tokens	1000000

The analyzed model identity and revision are stored authoritatively in icalens.json; model-card metadata is not used when loading the lens.

Usage

Analyze text end to end:

python

from icalens import ICALens

lens = ICALens.from_pretrained("REPOSITORY_ID")
result = lens.analyze("She deposited the check at the bank.", layer=0)

print(result.tokens)
print(result.scores)  # signed standard ICA scores
print(result.energy)  # per-token squared-score fractions

Or transform activations captured separately:

scores = lens.transform(activations, layer=0)

Externally captured activations must use the model revision, activation site, layer indexing, and preprocessing recorded in icalens.json.

Score definition

Signed scores are the centered, whitened activations followed by the learned orthogonal ICA rotation. No post-ICA source scaling is applied to v0.2 fits. For a token, component energy share is score² / sum(all component scores²).

Fitting

Layer	Components	Fitting tokens	FastICA iterations
0	512	1000000	50
1	512	1000000	50
2	512	1000000	50
3	512	1000000	50
4	512	1000000	50
5	512	1000000	50

Fitting provenance

json

{
  "activation_dataset": {
    "dtype": "bfloat16",
    "format": "icalens.activations",
    "format_version": 1,
    "manifest_sha256": "d77c1a6808227e5a7fc7e6487f20d27e0e325964b951eda1be6fb9b4427f3d88"
  },
  "candidate_tokens": 5317136,
  "context_length": 1024,
  "dataset": {
    "repo_id": "NeelNanda/pile-10k",
    "revision": "127bfedcd5047750df5ccf3a12979a47bfa0bafa",
    "split": "train"
  },
  "document_framing": {
    "evidence_note": "Pythia training documents are concatenated and separated by an EOD token.",
    "evidence_url": "https://github.com/EleutherAI/pythia#exploring-the-dataset",
    "included_in_fitting_samples": false,
    "policy_source": "project:model_framing.json",
    "registry_model_entry": "EleutherAI/pythia-70m",
    "registry_schema_version": 1,
    "registry_sha256": "c4929c65ecf42c17f99c3889a2057c89397da6d94f61d99b17ebecc32d1dde83",
    "requested": "auto",
    "strategy": "prepend-eos",
    "token": "<|endoftext|>",
    "token_id": 0
  },
  "fitting_tokens": 1000000,
  "sampling_seed": 0,
  "text_field": "text",
  "token_scope": "all"
}

Limitations

Component IDs are specific to a layer and fitted artifact.
Standard ICA scores are signed and are not probabilities.

Paper

ICA Lens: Interpreting Language Models Without Training Another Dictionary

Citation

bibtex

@article{liu2026icalens,
  title={ICA Lens: Interpreting Language Models Without Training Another Dictionary},
  author={Liu, Sida and Han, Feijiang},
  journal={arXiv preprint arXiv:2606.11722},
  year={2026}
}

sida/icalens-pythia-70m-pile10k