Remove workload trace configs from dataset card

README.md

@@ -32,14 +32,6 @@ configs:
     data_files:
       - split: train
         path: kernel_profiles/roofline_quadrant.parquet
-  - config_name: coding_agent_prompts
-    data_files:
-      - split: train
-        path: workload_traces/coding_agent_prompts.parquet
-  - config_name: osworld_trajectories
-    data_files:
-      - split: train
-        path: workload_traces/osworld_trajectories.parquet
   - config_name: predictions
     data_files:
       - split: train
@@ -259,32 +251,6 @@ dataset_info:
     splits:
       - name: train
         num_examples: 2163
-  - config_name: coding_agent_prompts
-    features:
-      - name: system
-        dtype: string
-      - name: user
-        dtype: string
-      - name: osl_tokens
-        dtype: int64
-    splits:
-      - name: train
-        num_examples: 500
-  - config_name: osworld_trajectories
-    features:
-      - name: session_id
-        dtype: string
-      - name: source
-        dtype: string
-      - name: num_turns
-        dtype: int64
-      - name: turns
-        dtype: string
-      - name: run
-        dtype: string
-    splits:
-      - name: train
-        num_examples: 60
   - config_name: predictions
     features:
       - name: hardware_config
@@ -374,11 +340,11 @@ dataset_info:
 
 # AgentPerfBench
 
-LLM inference benchmark: 3,392 serving runs, 148,077 per-kernel CUDA profiles, 4,715 latency predictions, and 560 workload traces across 9 models, 14 GPU configurations, and 2 serving engines (vLLM 0.19.0, SGLang 0.5.9). All models served in BF16 except gpt-oss, which uses mxfp4 for projection weights.
+LLM inference benchmark: 3,392 serving runs, 148,077 per-kernel CUDA profiles, and 4,715 latency predictions across 9 models, 14 GPU configurations, and 2 serving engines (vLLM 0.19.0, SGLang 0.5.9). All models served in BF16 except gpt-oss, which uses mxfp4 for projection weights.
 
 ## Dataset configurations
 
-Seven configurations covering serving benchmarks, kernel profiling, workload traces, and latency predictions.
+Five configurations covering serving benchmarks, kernel profiling, and latency predictions.
 
 ### trace_replay (3,147 rows)
 
@@ -400,16 +366,6 @@ Per-kernel CUDA profiling data from NCU (Nsight Compute). Individual kernel invo
 
 Operational intensity and achieved throughput per kernel, for roofline analysis. H100 reference hardware (989 peak TFLOPS, 3.35 TB/s HBM).
 
-### coding_agent_prompts (500 rows)
-
-System/user prompt pairs with output token counts from SWE-Bench coding agent sessions. Used to derive the trace_replay profiles.
-
-Full SWE-Bench and TerminalBench trajectory files are available separately; see the project repository.
-
-### osworld_trajectories (60 rows)
-
-Multi-turn OSWorld sessions with per-turn action/observation data (up to 30 turns per session). Used to derive the trace_replay profiles.
-
 ### predictions (4,715 rows)
 
 Predicted vs. measured latency for each serving configuration. Columns include ttft_pred/ttft_meas/ttft_err, tpot_pred/tpot_meas/tpot_err, e2el_pred/e2el_meas/e2el_err, plus cache-aware prediction metadata (cache_hit_rate, cache_aware_applied, multiturn_prediction_mode). Covers 14 hardware configs across all models and profiles.
@@ -496,8 +452,7 @@ from datasets import load_dataset
 
 # Serving benchmark results
 ds = load_dataset("agent-perf-bench/AgentPerfBench", "trace_replay")
-# or "distributional", "kernels_labeled", "roofline_quadrant",
-# "coding_agent_prompts", "osworld_trajectories", "predictions"
+# or "distributional", "kernels_labeled", "roofline_quadrant", "predictions"
 ```
 
 ## Benchmark methodology