Unverified50% confidenceTradeoffExact time
In context compression, average answer quality may remain stable, but a rare yet necessary piece of evidence may have quietly disappeared.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedResearch reveals that LLMs do not reliably weight and update based on evidence reliability like an ideal Bayesian agent, but instead exhibit behavioral patterns close to distributional skew, with their influence being less correlated with the true reliability of evidence than the Bayesian framework would predict.71% similarUnverifiedHigh-recall verification systems face four mutually conflicting constraints: reading less, maintaining high recall, not losing minority evidence, and not converting unchecked into nothing found.70% similarUnverifiedIn few-shot prompting, simply changing the order of examples can cause dramatic fluctuations in accuracy, ranging from near-random to near-optimal.69% similarUnverifiedChecking consistency through random sampling yields an inflated consistency rate because most samples are simple, masking the regions that are truly problematic.68% similarUnverifiedEven the most advanced current models have a notable gap in factual detail accuracy on summarization tasks.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949260API
curl https://kongchang.com/api/v1/knowledge/claims/949260MCP
get_claim(id=949260)