{ "band": 0.15, "band_sweep": { "0.05": { "hits": 64, "n": 78, "neutral_predicted": 12 }, "0.1": { "hits": 63, "n": 78, "neutral_predicted": 14 }, "0.15": { "hits": 63, "n": 78, "neutral_predicted": 14 }, "0.2": { "hits": 63, "n": 78, "neutral_predicted": 14 }, "0.3": { "hits": 63, "n": 78, "neutral_predicted": 14 }, "0.5": { "hits": 61, "n": 78, "neutral_predicted": 16 } }, "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": 0.8205128205128205, "confusion": [ [ 20, 1, 11 ], [ 0, 8, 0 ], [ 0, 2, 36 ] ], "directional_accuracy": 0.8, "directional_hits": 56, "directional_n": 70, "hits": 64, "macro_f1": 0.8194649519726919, "n": 78, "sign_agreement": 0.835820895522388, "sign_hits": 56, "sign_n": 67 }, "model (score through band \u00b10.15)": { "accuracy": 0.8076923076923077, "confusion": [ [ 20, 3, 9 ], [ 0, 8, 0 ], [ 0, 3, 35 ] ], "directional_accuracy": 0.7857142857142857, "directional_hits": 55, "directional_n": 70, "hits": 63, "macro_f1": 0.7833873443629541, "n": 78, "sign_agreement": 0.859375, "sign_hits": 55, "sign_n": 64 } }, "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_1946_transformer", "mcnemar": { "model_vs_deepseek": { "deepseek_only_right": 14, "model_only_right": 0, "p": 0.0001220703125 }, "model_vs_keyword": { "keyword_only_right": 1, "model_only_right": 41, "p": 1.9554136088117957e-11 } }, "model": { "final": false, "model_path": "/home/harwinder/Development/ML-NEWS/models/xau_transformer", "model_type": "two_head_transformer:microsoft/deberta-v3-small", "oos_from": "2026-09-18", "relevance_threshold": 0.2, "trained_at": "2026-09-27T19:46:15.509042+00:00", "training_examples": 10690 }, "model_latency_ms": { "max": 10.16, "mean": 8.463333333333333, "median": 8.425 }, "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": 7, "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": 5, "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": 1, "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": 2, "n": 7 }, "weak_usd": { "deepseek": 7, "deepseek_n": 7, "keyword": 0, "model": 6, "n": 7 } }, "spearman_model_vs_deepseek": 0.7803360556001124 }