README: describe reward schemes by behavior, not internal version names

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: gurusha01

README.md

@@ -152,10 +152,10 @@ per task lives in `configs/`:
 uv run rl --config configs/prime_rl_v9.2_titanic_tree.toml
 ```
 
-The reward is computed from the best score reached anywhere in the episode's tree.
-Three schemes are available (`reward_scheme` in the config): `v6_binary`,
-`v7_fixed_tier`, and `v9_percentile` (a rolling-percentile threshold; the most
-used). Rollout logs go to `$AIR_ROLLOUT_LOG_DIR` (default `./rollout_logs`).
+The reward is computed from the best score reached anywhere in the episode's tree;
+the reward shaping is selectable via `reward_scheme` in the config (binary,
+fixed-tier, or rolling-percentile). Rollout logs go to `$AIR_ROLLOUT_LOG_DIR`
+(default `./rollout_logs`).
 
 ## Reproducing the paper figures