{ "band": 0.15, "band_sweep": { "0.05": { "hits": 73, "n": 78, "neutral_predicted": 9 }, "0.1": { "hits": 72, "n": 78, "neutral_predicted": 10 }, "0.15": { "hits": 71, "n": 78, "neutral_predicted": 11 }, "0.2": { "hits": 70, "n": 78, "neutral_predicted": 13 }, "0.3": { "hits": 68, "n": 78, "neutral_predicted": 15 }, "0.5": { "hits": 66, "n": 78, "neutral_predicted": 18 } }, "blocks": { "DeepSeek live (xau through band \u00b10.15)": { "accuracy": 1.0, "confusion": [ [ 32, 0, 0 ], [ 0, 8, 0 ], [ 0, 0, 38 ] ], "directional_accuracy": 1.0, "directional_hits": 70, "directional_n": 70, "hits": 78, "macro_f1": 1.0, "n": 78, "sign_agreement": 1.0, "sign_hits": 70, "sign_n": 70 }, "keyword fallback (band \u00b10.15)": { "accuracy": 0.3076923076923077, "confusion": [ [ 6, 15, 11 ], [ 0, 8, 0 ], [ 12, 16, 10 ] ], "directional_accuracy": 0.22857142857142856, "directional_hits": 16, "directional_n": 70, "hits": 24, "macro_f1": 0.30646952758745044, "n": 78, "sign_agreement": 0.41025641025641024, "sign_hits": 16, "sign_n": 39 }, "model (native argmax label)": { "accuracy": 0.9230769230769231, "confusion": [ [ 28, 1, 3 ], [ 0, 8, 0 ], [ 1, 1, 36 ] ], "directional_accuracy": 0.9142857142857143, "directional_hits": 64, "directional_n": 70, "hits": 72, "macro_f1": 0.9139955369463566, "n": 78, "sign_agreement": 0.9411764705882353, "sign_hits": 64, "sign_n": 68 }, "model (score through band \u00b10.15)": { "accuracy": 0.9102564102564102, "confusion": [ [ 28, 1, 3 ], [ 0, 8, 0 ], [ 1, 2, 35 ] ], "directional_accuracy": 0.9, "directional_hits": 63, "directional_n": 70, "hits": 71, "macro_f1": 0.8937302272073627, "n": 78, "sign_agreement": 0.9402985074626866, "sign_hits": 63, "sign_n": 67 } }, "deepseek": { "batch_ms_max": null, "batch_ms_mean": null, "cache_hits": 78, "calls": 0, "failed_calls": 0, "headlines_sent": 0, "model": "deepseek-v4-flash", "per_headline_ms": null, "prompt_sha1": "d32e04c6a87d019f326f111b39c99b6dd8ce794d" }, "deepseek_scored": 78, "generated_at": "20260927_2315_transformer", "mcnemar": { "model_vs_deepseek": { "deepseek_only_right": 6, "model_only_right": 0, "p": 0.03125 }, "model_vs_keyword": { "keyword_only_right": 1, "model_only_right": 49, "p": 9.059419880941277e-14 } }, "model": { "final": true, "model_path": "/home/harwinder/Development/ML-NEWS/models/xau_transformer", "model_type": "two_head_transformer:microsoft/deberta-v3-small", "oos_from": "2026-09-27T23:15:30.403442+00:00", "relevance_threshold": 0.2, "trained_at": "2026-09-27T23:15:30.403442+00:00", "training_examples": 17245 }, "model_latency_ms": { "max": 10.06, "mean": 8.469871794871795, "median": 8.405000000000001 }, "n": 78, "per_category": { "distractors": { "deepseek": 8, "deepseek_n": 8, "keyword": 8, "model": 8, "n": 8 }, "dovish_fed": { "deepseek": 8, "deepseek_n": 8, "keyword": 7, "model": 8, "n": 8 }, "gold_direct": { "deepseek": 8, "deepseek_n": 8, "keyword": 4, "model": 6, "n": 8 }, "hawkish_fed": { "deepseek": 8, "deepseek_n": 8, "keyword": 2, "model": 7, "n": 8 }, "inverted_metrics": { "deepseek": 6, "deepseek_n": 6, "keyword": 0, "model": 6, "n": 6 }, "negation_traps": { "deepseek": 8, "deepseek_n": 8, "keyword": 2, "model": 7, "n": 8 }, "real_yields": { "deepseek": 6, "deepseek_n": 6, "keyword": 1, "model": 5, "n": 6 }, "risk_on": { "deepseek": 4, "deepseek_n": 4, "keyword": 0, "model": 4, "n": 4 }, "safe_haven": { "deepseek": 8, "deepseek_n": 8, "keyword": 0, "model": 8, "n": 8 }, "strong_usd": { "deepseek": 7, "deepseek_n": 7, "keyword": 0, "model": 7, "n": 7 }, "weak_usd": { "deepseek": 7, "deepseek_n": 7, "keyword": 0, "model": 6, "n": 7 } }, "spearman_model_vs_deepseek": 0.8597502720725484 }