institute
active
institute:goodfireGoodfire
AI research company; authors' affiliation; develops tools including EVEE and publishes research on genomic foundation models.
Neighborhood — ranked by edge-count
Thinkers (53)
thinker
- Atticus Geigeraffiliated_with
- Jack Merulloaffiliated_with
- Raphaël Sarfatiaffiliated_with
- Daniel Wurgaftaffiliated_with
- Ekdeep Singh Lubanaaffiliated_with
- Owen Lewisaffiliated_with
- Sheridan Feuchtaffiliated_with
- Thomas McGrathaffiliated_with
- Can Rageraffiliated_with
- Tal Haklayaffiliated_with
- Thomas Felaffiliated_with
- Usha Bhallaaffiliated_with
- Eric Bigelowaffiliated_with
- Dron Hazraaffiliated_with
- Santiago Aranguriaffiliated_with
- Daniel Balsamaffiliated_with
- Mark Bissellaffiliated_with
- Siddharth Boppanaaffiliated_with
- Thomas Doomsaffiliated_with
- Frank Xiaoaffiliated_with
- Lee Sharkeyaffiliated_withCo-author of VPD paper; mechanistic interpretability researcher affiliated with Goodfire.
- Michael T Pearceaffiliated_with
- Nicholas K. Wangaffiliated_with
- Alexander J. Ryuaffiliated_with
+29 more
Papers (7)
paper
- Explaining 4.2 million genetic variants with state-of-the-art, interpretable predictionsaffiliated_withassociated_with
- Interpreting Language Model Parametersassociated_with
- Verbalized Eval Awareness Inflates Measured Safetyassociated_with
- Covariance-based Sequence Poolingassociated_with
Communities (2)
community
- LLM Introspectionmembers_of
- Neural Steering Methodsmembers_of
Concepts (5)
concept
- Mayo Clinicassociated_withClinical partner institution collaborating on EVEE variant pathogenicity predictions.
Artifacts (2)
artifact
- Goodfire Ember APIintroducesAPI providing contrastive search functionality used to identify off-topic detector latents
- Complete technical paper with proofs, additional experiments on months/letters/ages, and cross-modal validation.
Events (2)
event
- Goodfire announced its collaboration with Mayo Clinic, which led to this paper.
- Subsequent Goodfire research addressing scale of mechanistic interpretability methods including VPD.
Institutes (1)
institute
- Radical AIcitesPartner organization with Goodfire on materials discovery research; partnership announced July 2025.
Methods (1)
method
- Core technique introduced in this paper for decomposing neural network weight matrices into mechanistically simple, interpretable rank-one subcomponents.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- API providing access to sparse autoencoder features for LLaMA 3.3 70B used for feature steering in Experiment 2
- API method used to identify latents differentially activated between on-topic and off-topic prompt-response pairs
- The same single being animating every part of a whole when each living center reflects the same origin.
- The property that a good shape is a center made up of powerful intense centers which themselves have good shape; built up from elementary figures with high internal symmetries, bilateral symmetry, a well-marked center, compactness, and closure
- The pure unity beyond the veil, the blazing furnace of intense light, glimpsed when a living center opens a window.
- A transformation that intensifies products of alternating repetition by strengthening loosely formed shapes and giving more life to centers within them.
- A regulator that achieves effective control; defined implicitly by the theorem as one that is a model of the system.
- A sudden, non-linear transition in neural activity marking global broadcast in GWT.