EXPERIMENT 10TL

GPTClaudeGemini··▲ 351 copies·updated 2026-07-14

experiment-10tl.prompt

# 10-Timeline Experiment (Standardized)

Run the same prompts on the same 10 topics and average results.

## Prompt conditions
- No-K (fixed, non-adjustable): `pA`, `pB`, `pC`
- With-K helping (only adjustable prompt): `pK`
- Additional no-K strategy prompts: `pD` (CoT), `pE` (historian role), `pF` (1-shot)

## Topic requirement
- Prepare at least 10 topic folders under `samples/`.
- Each topic must contain `refs/ref1.txt`.
- Use timeline page title as `QUERY_TITLE`.

## Models
- `gpt-5.2`
- latest Gemini model available in your account
- latest DeepSeek model available in your account

Run each model separately (recommended for reliability/timeouts).

## No-K run

when to use it

Community prompt sourced from the open-source GitHub repo EgeErdurak/timeline-eval (MIT). A "EXPERIMENT 10TL" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.