Liu Liu (@liuliu) asked why few people are dissecting Thinking Machines' downloadable Inkling 2 Small and listed what he found: Attention Residuals spanning 8 blocks of 6 layers in a 48-layer model, sigmoid gated attention plus a new per-token, per-he...
VQV Signal
OPEN SOURCE
WATCH
TECHNICAL
Inside Inkling 2 Small: attention residuals and a new attention sink
Liu Liu (@liuliu) asked why few people are dissecting Thinking Machines' downloadable Inkling 2 Small and listed what he found: Attention Residuals spanning 8 blocks of 6 layers in a 48-layer model, sigmoid gated attention plus a new per-token, per-he...
Liu Liu (@liuliu) asked why few people are dissecting Thinking Machines' downloadable Inkling 2 Small and listed what he found: Attention Residuals spanning 8 blocks of 6 layers in a 48-layer model, sigmoid gated attention plus a new per-token, per-head projected attention sink, and carry-overs such as relative...
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength 91%
Technical label WATCH
Public Interest 27
Category OPEN SOURCE
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 26
Novelty Interest Score 72
Consequence Score 18
Curiosity Score 0
Shareability Score 46
Topics
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.