Update README.md

README.md

@@ -28,7 +28,7 @@ written solution. The recommended checkpoint is **`opd-32b-bf16-step-225`** ("st
 
 | Benchmark | Result | Grader | Setting | Solutions |
 |---|---|---|---|---|
-| **IMO 2026** | **21 / 42 — Bronze medal** | Human expert (ex-IMO) | step-225, `high` budget | [submission.csv](https://huggingface.co/datasets/imo2026-challenge/chankhavu-imo-reasoning-traces/blob/main/imo2026-step225-budget-high-tournament/submission.csv) |
+| **IMO 2026** | **21 / 42 — Bronze medal** | [Human expert (ex-IMO)](https://github.com/fieldsmodelorg/AIMO-Proof-Pilot/tree/main/imo_2026_eval) | step-225, `high` budget | [solutions.csv](https://github.com/fieldsmodelorg/AIMO-Proof-Pilot/blob/main/imo_2026_eval/raw/imo2026-step225-budget-high-tournament_submission.csv) |
 | **IMO 2025** | **30 / 42 — Silver medal** | GPT-5.6-sol | step-225, `high` budget | [solutions.csv](https://github.com/fieldsmodelorg/AIMO-Proof-Pilot/blob/main/benchmarks/imo-2025/solutions.csv) |
 | **IMO-ProofBench-V2** | **66.4%** (mean **4.65 / 7**) | GPT-5.6-sol | step-225, `medium` budget | [solutions.csv](https://github.com/fieldsmodelorg/AIMO-Proof-Pilot/blob/main/benchmarks/imo-proofbench-v2/solutions.csv) |
 
@@ -131,7 +131,7 @@ live in [a separate HuggingFace repository](https://huggingface.co/fieldsmodelor
 step-225 at the `high` budget, graded per-problem by human experts (former IMO
 competitors) against broad IMO standards (no markscheme was available). The proofs are
 the tournament
-[`submission.csv`](https://huggingface.co/datasets/imo2026-challenge/chankhavu-imo-reasoning-traces/blob/main/imo2026-step225-budget-high-tournament/submission.csv);
+[`solutions.csv`](https://github.com/fieldsmodelorg/AIMO-Proof-Pilot/blob/main/imo_2026_eval/raw/imo2026-step225-budget-high-tournament_submission.csv);
 per-problem
 [`scores.csv`](https://github.com/fieldsmodelorg/AIMO-Proof-Pilot/blob/main/imo_2026_eval/scores.csv)
 and full commentary are in