docs: remove eval results references

README.md

@@ -67,8 +67,6 @@ Evaluation conditions: zero-shot, no chat template, FP32 evaluation, Apple Silic
 | MMLU (57-subject macro) | `acc` | 23.09% |
 | **Eight-task mean** | — | **30.00%** |
 
-The machine-readable results are in `.eval_results/base-100m.yaml` using the Hugging Face Hub Evaluation Results format. No `verifyToken` is included because this evaluation was not run through a verifiable Hugging Face evaluation job. The task datasets are recorded with their Hub IDs and revisions; Hub leaderboard aggregation additionally requires the dataset to be registered as a Benchmark with an `eval.yaml` task definition.
-
 ## Tokenizer
 
 - `Fleck-Tokenizer-1024`
@@ -138,6 +136,4 @@ The public bundle contains these files:
 - `config.json` — custom architecture configuration
 - `generation_config.json` — greedy generation defaults
 - `tokenizer_config.json` — tokenizer defaults and special-token mapping
-- `.eval_results/base-100m.yaml` — machine-readable evaluation results
-
 No training data, optimizer state, or other training outputs are included.