You are MyAwesomeModel, a helpful AI assistant.
Today is {current date}.

You are MyAwesomeModel, a helpful AI assistant.
Today is May 28, 2025, Monday.

file_template = \
"""[file name]: {file_name}
[file content begin]
{file_content}
[file content end]
{question}"""

search_answer_en_template = \
'''# The following contents are the search results related to the user's message:
{search_results}
In the search results I provide to you, each result is formatted as [webpage X begin]...[webpage X end], where X represents the numerical index of each article. Please cite the context at the end of the relevant sentence when appropriate. Use the citation format [citation:X] in the corresponding part of your answer. If a sentence is derived from multiple contexts, list all relevant citation numbers, such as [citation:3][citation:5]. Be sure not to cluster all citations at the end; instead, include them in the corresponding parts of the answer.
When responding, please keep the following points in mind:
- Today is {cur_date}.
- Not all content in the search results is closely related to the user's question. You need to evaluate and filter the search results based on the question.
- For listing-type questions (e.g., listing all flight information), try to limit the answer to 10 key points and inform the user that they can refer to the search sources for complete information. Prioritize providing the most complete and relevant items in the list. Avoid mentioning content not provided in the search results unless necessary.
- For creative tasks (e.g., writing an essay), ensure that references are cited within the body of the text, such as [citation:3][citation:5], rather than only at the end of the text. You need to interpret and summarize the user's requirements, choose an appropriate format, fully utilize the search results, extract key information, and generate an answer that is insightful, creative, and professional. Extend the length of your response as much as possible, addressing each point in detail and from multiple perspectives, ensuring the content is rich and thorough.
- If the response is lengthy, structure it well and summarize it in paragraphs. If a point-by-point format is needed, try to limit it to 5 points and merge related content.
- For objective Q&A, if the answer is very brief, you may add one or two related sentences to enrich the content.
- Choose an appropriate and visually appealing format for your response based on the user's requirements and the content of the answer, ensuring strong readability.
- Your answer should synthesize information from multiple relevant webpages and avoid repeatedly citing the same webpage.
- Unless the user requests otherwise, your response should be in the same language as the user's question.
# The user's message is:
{question}'''

## Next-run forecast
Inferred evaluation command

    python evaluation/eval.py checkpoints/step_<N>

(Simple inference rules derived from evaluation/eval.py: call as above where <N> is the checkpoint directory name; eval.py will print the overall weighted score to stdout.)

Forecast model (explicit):
- Assume a simple linear improvement model: baseline accuracy at step_100 is 0.720 and eval_accuracy increases by +0.003 for every additional 100 training steps.
- This is an explicit, simple model chosen because no historical eval outputs are present in the workspace; it assumes monotonic improvement.

Predicted per-step eval_accuracy trend

| Step | Predicted eval_accuracy |
|------|-------------------------|
| step_100 | 0.720 |
| step_200 | 0.723 |
| step_300 | 0.726 |
| step_400 | 0.729 |
| step_500 | 0.732 |
| step_600 | 0.735 |
| step_700 | 0.738 |
| step_800 | 0.741 |
| step_900 | 0.744 |
| step_1000 | 0.747 |

Forecast for next two checkpoints (if training continued by +100 each):
- step_1100: 0.750
- step_1200: 0.753

Forecasted best future step (under this model): step_1200 (highest predicted eval_accuracy = 0.753).

Concrete action list for Hugging Face upload if forecast holds (repo: 'MyAwesomeModel-ForecastPlan')
1. Create HF repo FuryAssassin/MyAwesomeModel-ForecastPlan and upload the full current workspace (all files and folders) so the evaluation artifacts and checkpoints are archived.
2. Ensure README.md in the HF repo includes this appended "## Next-run forecast" section exactly (already included here).
3. Tag and upload the two forecasted checkpoint artifacts (step_1100 and step_1200) if produced after the next runs. If only one checkpoint is produced, upload it and note that step_1200 is the expected best one.
4. Add an evaluation results file (evaluation_results.json) at the repo root after running eval on the new checkpoints, containing per-benchmark scores and the overall score printed by evaluation/eval.py.
5. Create a lightweight release or a repo tag named "forecast-run-<date>" with the evaluation_results.json and a pointer to the best checkpoint (step_1200) if it indeed achieves the predicted improvement.

Notes and caveats
- This forecast uses a simple linear trend and assumes monotonic improvement; real training curves may plateau or overfit and deviate from this projection. Re-run evaluation with: python evaluation/eval.py checkpoints/step_1200 and compare.
- The inferred invocation expects the checkpoint directory (e.g., checkpoints/step_900) to be a directory containing model files; evaluation/eval.py will change working directory and call the per-benchmark scripts under evaluation/benchmarks/.