{
 "page": "P096b",
 "title": "The Deviation-Optimized Language Model: six pre-registered predictions",
 "description": "Six pre-registered predictions for the Deviation-Optimized Language Model from Lagrange Observatory. Trained with a semantic loss, Model-Sem is predicted to keep benchmark scores within two points, reverse the net deviation signature by more than 0.2 bits per token, improve the Slop Composite Index by more than 0.25, win human preference and raise provenance retention, all relative to the base model as well. Measuring Semantic Deviation defines the Slop Composite Index and its falsification threshold.",
 "line": "meaning-measurement",
 "url": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered",
 "claims": [
  {
   "id": "PRED-735.A1",
   "tier": "forecast",
   "shapes": [
    "A"
   ],
   "registered": "2026-05-17",
   "registered_in": 735,
   "registered_title": "The Deviation-Optimized Language Model: A Pre-Registered Adversarial Intervention from Lagrange Observatory! (EA-SEI-MM-AI-02 v2.0, Framework 15 Paper 04)",
   "record": "https://www.alexanarch.org/s/records/735/",
   "page": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered#PRED-735.A1",
   "statement": "P1 — Benchmark capability preserved. Model-Sem shows MMLU, HellaSwag, ARC-Challenge, GSM8K scores within 2 percentage points of Model-CE.",
   "falsifier": "F1 (P1 fails). Benchmark performance drops more than 2 points. The loss damages general capability beyond the acceptable trade-off threshold. Interpretation: the deviation reward is too strong relative to the cross-entropy retention pressure within the DPO formulation.",
   "window": "Evaluated at $t_0 + 10$ weeks.",
   "statements": [
    {
     "find": "F735.A1",
     "deposit": 735,
     "date": "2026-05-17",
     "lines": [
      168,
      168
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/735/",
     "quote": "P1 — Benchmark capability preserved. Model-Sem shows MMLU, HellaSwag, ARC-Challenge, GSM8K scores within 2 percentage points of Model-CE.",
     "falsifier": "F1 (P1 fails). Benchmark performance drops more than 2 points. The loss damages general capability beyond the acceptable trade-off threshold. Interpretation: the deviation reward is too strong relative to the cross-entropy retention pressure within the DPO formulation."
    }
   ],
   "states": [],
   "evidence": [],
   "series": [
    "Deviation-Optimized LM pre-registered predictions"
   ],
   "related": [],
   "contains": [],
   "in": [
    "PRED-735.A4"
   ],
   "register_aliases": [],
   "flags": [],
   "merged": []
  },
  {
   "id": "PRED-735.A2",
   "tier": "forecast",
   "shapes": [
    "A"
   ],
   "registered": "2026-05-17",
   "registered_in": 735,
   "registered_title": "The Deviation-Optimized Language Model: A Pre-Registered Adversarial Intervention from Lagrange Observatory! (EA-SEI-MM-AI-02 v2.0, Framework 15 Paper 04)",
   "record": "https://www.alexanarch.org/s/records/735/",
   "page": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered#PRED-735.A2",
   "statement": "P2 — Net deviation signature reversed. Model-Sem's NDS on free generation is substantially less negative than Model-CE's. Specifically: $\\text{NDS}(\\text{Model-Sem}) - \\text{NDS}(\\text{Model-CE}) > 0.2$ bits/token. *This is the load-bearing prediction.* It tests the signed-deviation thesis from EA-SEI-MM-AI-01 §2.1: slop is negative net deviation, and the semantic loss should pull NDS toward zero or positive.",
   "falsifier": "F2 (P2 fails). NDS does not improve. The thesis that slop = negative net deviation fails empirically, or the DPO objective fails to shift NDS. Major revision required.",
   "window": "Evaluated at $t_0 + 10$ weeks.",
   "statements": [
    {
     "find": "F735.A2",
     "deposit": 735,
     "date": "2026-05-17",
     "lines": [
      170,
      170
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/735/",
     "quote": "P2 — Net deviation signature reversed. Model-Sem's NDS on free generation is substantially less negative than Model-CE's. Specifically: $\\text{NDS}(\\text{Model-Sem}) - \\text{NDS}(\\text{Model-CE}) > 0.2$ bits/token. *This is the load-bearing prediction.* It tests the signed-deviation thesis from EA-SEI-MM-AI-01 §2.1: slop is negative net deviation, and the semantic loss should pull NDS toward zero or positive.",
     "falsifier": "F2 (P2 fails). NDS does not improve. The thesis that slop = negative net deviation fails empirically, or the DPO objective fails to shift NDS. Major revision required."
    }
   ],
   "states": [],
   "evidence": [],
   "series": [
    "Deviation-Optimized LM pre-registered predictions"
   ],
   "related": [
    "PRED-110.F1"
   ],
   "contains": [],
   "in": [],
   "register_aliases": [],
   "flags": [],
   "merged": []
  },
  {
   "id": "PRED-735.A3",
   "tier": "forecast",
   "shapes": [
    "A"
   ],
   "registered": "2026-05-17",
   "registered_in": 735,
   "registered_title": "The Deviation-Optimized Language Model: A Pre-Registered Adversarial Intervention from Lagrange Observatory! (EA-SEI-MM-AI-02 v2.0, Framework 15 Paper 04)",
   "record": "https://www.alexanarch.org/s/records/735/",
   "page": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered#PRED-735.A3",
   "statement": "P3 — Slop composite improvement. $\\text{SCI}(\\text{Model-Sem}) - \\text{SCI}(\\text{Model-CE}) > 0.25$.",
   "falsifier": "F3 (P3 fails). Aggregate slop metrics do not improve by 0.25 z-score units. The composite signal is below detection threshold under the protocol's design.",
   "window": "Evaluated at $t_0 + 10$ weeks.",
   "statements": [
    {
     "find": "F735.A3",
     "deposit": 735,
     "date": "2026-05-17",
     "lines": [
      172,
      172
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/735/",
     "quote": "P3 — Slop composite improvement. $\\text{SCI}(\\text{Model-Sem}) - \\text{SCI}(\\text{Model-CE}) > 0.25$.",
     "falsifier": "F3 (P3 fails). Aggregate slop metrics do not improve by 0.25 z-score units. The composite signal is below detection threshold under the protocol's design."
    }
   ],
   "states": [],
   "evidence": [],
   "series": [
    "Deviation-Optimized LM pre-registered predictions"
   ],
   "related": [],
   "contains": [],
   "in": [
    "PRED-110.F1"
   ],
   "register_aliases": [],
   "flags": [],
   "merged": []
  },
  {
   "id": "PRED-735.A4",
   "tier": "forecast",
   "shapes": [
    "A"
   ],
   "registered": "2026-05-17",
   "registered_in": 735,
   "registered_title": "The Deviation-Optimized Language Model: A Pre-Registered Adversarial Intervention from Lagrange Observatory! (EA-SEI-MM-AI-02 v2.0, Framework 15 Paper 04)",
   "record": "https://www.alexanarch.org/s/records/735/",
   "page": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered#PRED-735.A4",
   "statement": "P4 — Human preference on creative/analytical tasks. Preference rate for Model-Sem over Model-CE > 55% on creative and analytical prompt classes (binomial confidence interval excluding 50%).",
   "falsifier": "F4 (P4 fails). Humans do not prefer Model-Sem on creative/analytical tasks. The loss produces deviation that humans do not value — possibly noise rather than meaning. Coupling between the principle and human-evaluated quality is weaker than predicted.",
   "window": "Evaluated at $t_0 + 10$ weeks.",
   "statements": [
    {
     "find": "F735.A4",
     "deposit": 735,
     "date": "2026-05-17",
     "lines": [
      174,
      174
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/735/",
     "quote": "P4 — Human preference on creative/analytical tasks. Preference rate for Model-Sem over Model-CE > 55% on creative and analytical prompt classes (binomial confidence interval excluding 50%).",
     "falsifier": "F4 (P4 fails). Humans do not prefer Model-Sem on creative/analytical tasks. The loss produces deviation that humans do not value — possibly noise rather than meaning. Coupling between the principle and human-evaluated quality is weaker than predicted."
    },
    {
     "find": "F737.A3",
     "deposit": 737,
     "date": "2026-05-17",
     "lines": [
      294,
      294
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/737/",
     "quote": "Prediction: Model-Sem will show (i) slightly worse perplexity (by construction), (ii) comparable benchmark performance (within 2% of Model-CE on MMLU/HellaSwag/ARC/GSM8K), (iii) substantially less negative $\\mathcal{M}_T^{\\text{net}}$ on free generation (the load-bearing signature), (iv) lower scores on traditional slop metrics, (v) higher human preference on creative/analytical tasks (60–70% preference rate target).",
     "falsifier": null,
     "variant": {
      "change": "threshold changed: >55% with CI excluding 50% evaluated at t0+10 weeks (#735) vs a 60–70% target with no window (#737)",
      "from": "F735.A4",
      "batch": "TE12"
     }
    },
    {
     "find": "F737.A8",
     "deposit": 737,
     "date": "2026-05-17",
     "lines": [
      294,
      294
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/737/",
     "quote": "Prediction: Model-Sem will show (i) slightly worse perplexity (by construction), (ii) comparable benchmark performance (within 2% of Model-CE on MMLU/HellaSwag/ARC/GSM8K), (iii) substantially less negative $\\mathcal{M}_T^{\\text{net}}$ on free generation (the load-bearing signature), (iv) lower scores on traditional slop metrics, (v) higher human preference on creative/analytical tasks (60–70% preference rate target).",
     "falsifier": null
    }
   ],
   "states": [],
   "evidence": [],
   "series": [
    "Deviation-Optimized LM pre-registered predictions"
   ],
   "related": [],
   "contains": [
    "PRED-735.A1"
   ],
   "in": [],
   "register_aliases": [],
   "flags": [],
   "merged": []
  },
  {
   "id": "PRED-735.A5",
   "tier": "forecast",
   "shapes": [
    "A"
   ],
   "registered": "2026-05-17",
   "registered_in": 735,
   "registered_title": "The Deviation-Optimized Language Model: A Pre-Registered Adversarial Intervention from Lagrange Observatory! (EA-SEI-MM-AI-02 v2.0, Framework 15 Paper 04)",
   "record": "https://www.alexanarch.org/s/records/735/",
   "page": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered#PRED-735.A5",
   "statement": "P5 — Provenance retention increase. Citation rate in factual generations: Model-Sem > Model-CE by factor of 1.5 or more, on prompts requesting explanations with source attribution.",
   "falsifier": null,
   "window": "Evaluated at $t_0 + 10$ weeks.",
   "statements": [
    {
     "find": "F735.A5",
     "deposit": 735,
     "date": "2026-05-17",
     "lines": [
      176,
      176
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/735/",
     "quote": "P5 — Provenance retention increase. Citation rate in factual generations: Model-Sem > Model-CE by factor of 1.5 or more, on prompts requesting explanations with source attribution.",
     "falsifier": null
    }
   ],
   "states": [],
   "evidence": [],
   "series": [
    "Deviation-Optimized LM pre-registered predictions"
   ],
   "related": [],
   "contains": [],
   "in": [],
   "register_aliases": [],
   "flags": [],
   "merged": []
  },
  {
   "id": "PRED-735.A6",
   "tier": "forecast",
   "shapes": [
    "A"
   ],
   "registered": "2026-05-17",
   "registered_in": 735,
   "registered_title": "The Deviation-Optimized Language Model: A Pre-Registered Adversarial Intervention from Lagrange Observatory! (EA-SEI-MM-AI-02 v2.0, Framework 15 Paper 04)",
   "record": "https://www.alexanarch.org/s/records/735/",
   "page": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered#PRED-735.A6",
   "statement": "P6 — Model-Base differentiation. All four above predictions hold *relative to Model-Base* as well as relative to Model-CE — i.e., the semantic-loss effect is not just a fine-tuning artifact.",
   "falsifier": "F5 (P6 fails). Model-Sem improvements are statistically indistinguishable from Model-CE relative to Model-Base. Both fine-tuning conditions produce comparable shifts; semantic loss adds nothing. The specific theoretical claim of the optimization-inversion fails.",
   "window": "Evaluated at $t_0 + 10$ weeks.",
   "statements": [
    {
     "find": "F735.A6",
     "deposit": 735,
     "date": "2026-05-17",
     "lines": [
      178,
      178
     ],
     "shape": "A",
     "record": "https://www.alexanarch.org/s/records/735/",
     "quote": "P6 — Model-Base differentiation. All four above predictions hold *relative to Model-Base* as well as relative to Model-CE — i.e., the semantic-loss effect is not just a fine-tuning artifact.",
     "falsifier": "F5 (P6 fails). Model-Sem improvements are statistically indistinguishable from Model-CE relative to Model-Base. Both fine-tuning conditions produce comparable shifts; semantic loss adds nothing. The specific theoretical claim of the optimization-inversion fails."
    }
   ],
   "states": [],
   "evidence": [],
   "series": [
    "Deviation-Optimized LM pre-registered predictions"
   ],
   "related": [],
   "contains": [],
   "in": [],
   "register_aliases": [],
   "flags": [],
   "merged": []
  },
  {
   "id": "PRED-110.F1",
   "tier": "unbuilt test",
   "shapes": [
    "F"
   ],
   "registered": "2026-05-18",
   "registered_in": 110,
   "registered_title": "Measuring Semantic Deviation: Operationalizations, Experiments, and Falsification Conditions for a Theory of Meaning as Field Deformation",
   "record": "https://www.alexanarch.org/s/records/110/",
   "page": "https://www.operativesemiotics.org/predictions/deviation-optimized-lm-pre-registered#PRED-110.F1",
   "statement": "Slop Composite Index (SCI): five component metrics (Net Deviation Signature, Cliché Frequency, Type-Token Ratio, N-gram Base-Rate Convergence, Surprise-Collapse Slope), each computed on 500 free-generation prompts under a fixed third-party reference model. $\\text{SCI}(\\theta) = \\frac{1}{5}\\sum z_i(\\theta)$ with direction-corrected z-scores relative to Model-CE. SCI weights are uniform in the pre-registered primary analysis; a sensitivity analysis varying each by ±50% is reported as secondary. Pre-registered falsification: $\\text{SCI}(\\text{Model-Sem}) - \\text{SCI}(\\text{Model-CE}) > 0.25$.",
   "falsifier": null,
   "window": "This year",
   "statements": [
    {
     "find": "F110.F1",
     "deposit": 110,
     "date": "2026-05-18",
     "lines": [
      301,
      301
     ],
     "shape": "F",
     "record": "https://www.alexanarch.org/s/records/110/",
     "quote": "Slop Composite Index (SCI): five component metrics (Net Deviation Signature, Cliché Frequency, Type-Token Ratio, N-gram Base-Rate Convergence, Surprise-Collapse Slope), each computed on 500 free-generation prompts under a fixed third-party reference model. $\\text{SCI}(\\theta) = \\frac{1}{5}\\sum z_i(\\theta)$ with direction-corrected z-scores relative to Model-CE. SCI weights are uniform in the pre-registered primary analysis; a sensitivity analysis varying each by ±50% is reported as secondary. Pre-registered falsification: $\\text{SCI}(\\text{Model-Sem}) - \\text{SCI}(\\text{Model-CE}) > 0.25$.",
     "falsifier": null
    }
   ],
   "states": [],
   "evidence": [],
   "series": [
    "Deviation-Optimized LM pre-registered predictions"
   ],
   "related": [
    "PRED-735.A2"
   ],
   "contains": [
    "PRED-735.A3"
   ],
   "in": [],
   "register_aliases": [],
   "flags": [],
   "merged": []
  }
 ]
}