VQV Signal
According to a post on X by @WesRoth, MIT researchers have described a "telepathy" method that lets AI models share their internal states directly instead of writing text to each other. In one test, Google's Gemma 3 4B passed what it saw in images to IBM's text-only Granite 3.3 8B. Granite then averaged 41.6% on five visual question benchmarks, compared with 21.1% when it read a text description written by Gemma.
According to a post on X by @WesRoth, MIT researchers have described a "telepathy" method that lets AI models share their internal states directly instead of writing text to each other. In one test, Google's Gemma 3 4B passed what it saw in images to IBM's text-only Granite 3.3 8B. Granite then averaged 41.6% on five visual question benchmarks, compared with 21.1% when it read a text description written by Gemma.
AI-assisted summary based on the listed source.
MIT researchers' "telepathy" method lets AI models share their internal states directly rather than writing text to each other. In one test, Google's Gemma 3 4B passed what it saw in images to IBM's text-only Granite 3.3 8B, which then averaged 41.6% on five visual question benchmarks versus 21.1% when reading...
If these results hold, the approach could change how builders design systems where one AI model hands work to another. The figures come from the post's summary of the MIT work, so the original research should be checked for full methods and scope. Watch for the full paper and whether other models and tasks show similar gains.
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
VQV surfaced this signal because it is recent, relevant to AI Search, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.
