Can Attention Handle Disagreement?
Channel:
Subscribers:
846
Published on ● Video Link: https://www.youtube.com/watch?v=t2J6o8LJUeM
Language models excel at agreement and elaboration because the attention mechanism focuses on related input. Disagreement requires stepping away from context, which attention doesn’t naturally support. As a result, LLMs tend to amplify rather than oppose what’s already being said.
Other Videos By sakkharin
2025-04-02 | Programming Languages Are Like DNA |
2025-04-02 | Thank you for my trouble, VJ |
2025-04-01 | 0320 15 copy copy 9:16 |
2025-04-01 | simple marble toy demo created using Matter.js and p5.js |
2025-04-01 | the Carnian Pluvial Episode Nature Ambient Series |
2025-04-01 | N-GRAM EXPERIMENT 8bit like pops |
2025-04-01 | Is the C Array Just an Illusion? |
2025-03-31 | Capital: On Interest and Love |
2025-03-31 | Attention music to enrich your life. 0320 15 |
2025-03-31 | BGM 0315 5 |
2025-03-31 | Can Attention Handle Disagreement? |
2025-03-30 | MMT, Keynes, and the Tax Puzzle |
2025-03-30 | A formula that symbolizes the essence of LLMs and information theory |
2025-03-30 | sakkharin is live |
2025-03-28 | MagicaVoxel & Three.js: File Compatibility Explained shorts |
2025-03-28 | Sora Interlude. 0322 3 No Flubs - Surf Ninja |
2025-03-28 | LOFAS |
2025-03-28 | color theory history |
2025-03-28 | interlude 0320 5 |
2025-03-28 | WHAT IF 0328 1 |
2025-03-27 | GLSL 2*(1.0+sin(x))*pow(cos(x-atan2(sin(t)*4,1)*2.0),2.) |