Unverified50% confidenceFactExact time
Early language models from the GPT-2 era tended to 'forget' earlier context in long texts, resulting in contradictory character behavior or plot discontinuities.
1
Sources
50%
Confidence
Long-term
Relevance
10/10/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedResearch shows that large language models exhibit a significant "Lost in the Middle" phenomenon when processing long texts, retaining information from the beginning and end of the input text well, but showing a marked decline in attention to information in the middle.67% similarVerifiedLarge language models are inherently stateless, and remembering conversation history relies on concatenating the historical conversation text and feeding it into the model together.66% similarUnverifiedCurrent mainstream large language models (such as GPT, Claude, etc.) adopt a session-based interaction mode; the context window has expanded from an initial 4K tokens to 128K or even longer, but it remains essentially a single-conversation memory mechanism, and implicitly accumulated knowledge is lost once a new session begins.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/993159API
curl https://kongchang.com/api/v1/knowledge/claims/993159MCP
get_claim(id=993159)