Unsupervised Method Measures Whether LLM Messages Are Received
sanmikoyejo · x · 2026-10-31
Researchers propose an unsupervised measurement system to tell whether a message was received between LLMs, without human labels.
- Mechanism: A transmitter LLM writes a prompt to send a signal about a construct to a receiver LLM; the receiver tries to reconstruct the prompt, and the degree of reconstruction is the success measure.
- Details: The transmitter writes prompt p and labels items x with measures m; the receiver sees (x, m), writes a prompt p' that would have generated this labeling scheme, and labels m'; m and m' are compared.
- Findings: The measure is more useful for optimizers than standard unsupervised metrics like self-consistency and reflective LLM judgments, and it closely tracks human labels where ground truth exists.
More from Research
- Lab grows neurons on chips to slash video model costs 8x vs Wan 2.2 at $0.05 per clip — PeterDiamandis · 2026-10-31
- Neural Thickets: adding noise to LLM weights boosts reasoning for free — it's just sampling in disguise — kastnerkyle · 2026-10-31
- Study extracts chromatin QTLs from cell-free epigenomic liquid biopsy data — anshulkundaje · 2026-10-31
- Ropedia releases DONIC: one policy learns all two-handed robot manipulation, zero fine-tuning — liuziwei7 · 2026-10-31
- Real human brain cells are being used for AI, and Reddit is buzzing — EmotionalGuess9229 · 2026-10-31
- LLM judges harbor systematic errors that silently corrupt model rankings and training pipelines — burkov · 2026-10-31