Watts-per-Intelligence Part II: Algorithmic Catalysis
2026-04-24T08:51:42Z•83df3b37bcdbf9725495f2141b5feefc1003628126af86d43cb49b114b7201dc
2-D polar codingCramér–Rao boundDiP-SDLLM-in-the-loopLandauer erasureMDP convolutional codesMLEPMI feedbackURLLCalgorithmic catalysisalgorithmic mutual informationarXivchannel estimationcoding theorydistributed inferenceedge computinginformation theorylocally recoverable codesmassive MIMOmatrix completionmaximally recoverable codespolar codesrobust beamforming','posterior Cramér–Rao bound','S-procedure','speculative decodingthermodynamics of computation
What happened
Collection of arXiv papers (2026-04-24) across information theory, wireless communications, coding theory, and ML-for-communications. Highlights: a thermodynamic theory of "algorithmic catalysis" that bounds class-specific computational speed-ups by algorithmic mutual information and shows a minimum Landauer erasure cost; DiP‑SD, a distributed pipelined speculative decoding scheme for multi-user edge LLM inference (up to 17.89x throughput vs autoregressive decoding); maximum-likelihood channel estimation from PMI-only feedback in FDD massive MIMO with Cramér–Rao analysis and sharp excess-risk/
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_math_it
- Record identifier
- 83df3b37bcdbf9725495f2141b5feefc1003628126af86d43cb49b114b7201dc
- Enrichment time
- 2026-04-24T08:51:42Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.