Home▸AI & Machine Learning▸Lost in the Middle: Long-Context Recall Curve (2D)

Lost in the Middle: Long-Context Recall Curve (2D)

Interactive 2D simulator of the 'lost in the middle' effect: place a needle fact at any depth in a long context window, run Monte Carlo recall trials, and watch the empirical U-shaped recall curve emerge against the theoretical model in a layered dual-panel chart.

AI & Machine Learning2DModerate60 FPS📱 Mobile-adapted⇄ 3D version

Long-context language models don't treat every position in a huge context window equally. This 2D companion renders the context window as a flat row of chunks — height and color encode the theoretical probability that a fact placed there would be recalled — with a needle marker you can drag to any depth to run Monte Carlo recall trials. A second panel below plots the theoretical recall curve R(d) alongside the live empirical curve building up from your trials. Tune context length and simulated model quality to watch the classic U-shaped "lost in the middle" curve sharpen or flatten, and use Auto-Sweep to build the empirical curve automatically across every depth.

⚙ Under the hood

Drag a needle fact to any depth inside a long context window and run Monte Carlo recall trials in a layered dual-panel 2D chart — a chunk row plus a live theoretical-vs-empirical recall curve — to watch the classic U-shaped 'lost in the middle' effect emerge as you tune context length and simulated model quality.

llmlong-contextretrievalattentionragnlp

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install