Skip to content
OpenTrain AIFor AI Companies

Eliciting Latent Predictions from Transformers with the Tuned Lens

Published Mar 1, 2023
arXiv PDF
Researcher verdict
Starting point
Use as implementation starting point
Benchmark evidence
Missing
Not verified yet
Time to first repro
A few hours
Fast first run
Risk flags
0
None detected

Results and benchmarks

Freshness tier: cold
Eliciting Latent Predictions from Transformers with the Tuned Lens presents a transformer method.

Implementation

Best maintained implementation now

Recommended
Confidence: High
Reproducibility: Strong
neelnanda-io/transformerlens

A library for mechanistic interpretability of GPT-style language models

3,818 stars · 678 forks · Last push Aug 24, 2026 · MIT license

  • License
  • CI
  • Dependencies
  • Docker

Official implementation from Papers with Code · Repository link is mentioned in the paper metadata · Community adoption signal (3818 stars)

Why this implementation
Confidence: high

neelnanda-io/transformerlens is the strongest maintained implementation based on ranking signals. CI workflows are present. License is declared (MIT).

Open neelnanda-io/transformerlens
Reproduction risks
  • No repository-level red flags were detected, but paper-specific preprocessing and hyperparameter details may still be under-specified.
  • Selected neelnanda-io/transformerlens as the strongest maintained implementation for new work.
  • Includes CI workflow signals.
  • Includes dependency/environment manifest signals.
  • Repository activity is within the last 24 months.

Compare implementation paths

Compare maintenance quality, reproducibility coverage, and evidence confidence before choosing a reproduction baseline.

neelnanda-io/transformerlens
best maintained
Maintenance
Active
Confidence
High
Reproducibility
Strong
Stars
3,818
Last push
Aug 24, 2026 (1d)

Official implementation from Papers with Code · Repository link is mentioned in the paper metadata

  • No Docker setup
alignmentresearch/tuned-lens
historical official
Maintenance
Stale
Confidence
High
Reproducibility
Strong
Stars
610
Last push
Aug 7, 2025 (383d)

Official implementation from Papers with Code · Repository link is mentioned in the paper metadata

  • No push in 12+ months
EleutherAI/pythia
alternative
Maintenance
Stale risk
Confidence
Low
Reproducibility
Moderate
Stars
2,909
Last push
Nov 15, 2025 (283d)

Matched via arXiv identifier search · Community adoption signal (2909 stars)

  • No CI pipeline detected
  • No tagged releases
  • No Docker setup

Reproduction readiness

Time to first repro: hours
Last checked: Aug 24, 2026

Ready to run

Ready to reproduce

  • Clone neelnanda-io/transformerlens and install dependencies from pyproject.toml.
  • CI pipeline detected, so automated tests are in place.
  • Last updated 1 days ago.
Open neelnanda-io/transformerlens

Quick start

git clone https://github.com/neelnanda-io/transformerlens.git
pip install -e .

Repositories and ecosystem

No additional verified repositories beyond the primary recommendation.

Hugging Face artifacts

No trustworthy direct or curated related Hugging Face artifacts were found yet. Use targeted searches to quickly locate candidate models, datasets, and demos.

Search models Search datasets Search spaces

Tip: start with models, then check datasets and spaces if you need evaluation data or demos.

Research context

Tasks

None detected

Methods

Transformer

Domains

None detected

Evaluation and human feedback data

Open this paper in HFEPX to review benchmark signals, evaluation modes, and human-feedback protocol context.

Open in HFEPX
Explore similar papers

Jump to Paper2Code search queries derived from this paper's research context.

Transformer

Data includes links from Papers with Code ( CC-BY-SA-4.0 ).