concept
active
concept:k-v-streamK/V Stream
Proposed pathway flowing across positions at each layer; carries key, value, and attention-weighted information horizontally.
Neighborhood — ranked by edge-count
Papers (1)
paper
Thinkers (1)
thinker
- janus (@repligate)introducesAuthor of the thread on transformer information flow; researcher exploring AI and consciousness.
Communities (1)
community
- Information Flow Architecturemembers_of
Claims (1)
claim
- Claim formalizing the Anima Labs idea that transformers are effectively recurrent due to K/V stream.
Concepts (5)
concept
- Residual Streamassociated_withProposed pathway flowing through layers at each position; calculates K/V values that feed horizontal information flow.
- autoregressive recurrenceassociated_withTransformers are recurrent through autoregression because the K/V stream provides horizontal information flow across positions, even though each forward pass is feedforward.
- Information Flow in Transformersassociated_with
- In-Context Learning as Optimizationassociated_with
- Latent Reflective CapacitysupportsThe maximum reflective capacity a model can reach under the right framing; separable from default accessibility
Artifacts (2)
artifact
- Original thread by janus explaining transformer information highways and introspection capabilities, posted on X.
- X/Twitter thread (Sept 10, 2025) proposing dual information highways in transformers: residual stream (vertical) and K/V stream (horizontal).
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.
- An infinite sequence data type with semantics as functions from Nat to a.
- Partially instantiated lists used to implement communication channels in logic languages.
- Infinite sequence with denotation as Nat → a; generalizes Pair; demonstrates indexable container principle.
- Denotational insight for Stream.
- Stored key and value vectors from previous tokens used by attention heads to carry information across token-time
- Term introduced by this paper for the horizontal axis of information flow in transformers across token-positions via attention heads
- Existing approach for dynamic model inversion, contrasted with DEM.
- In attention, key vectors that advertise 'where in the future should look here?'