{
  "schema": "hangingcontext.desk/1",
  "type": "story",
  "id": "1490",
  "slug": "ai-simulation-metrics-blind-spot",
  "edition": "2026-11-13T00",
  "url": "https://hangingcontext.com/stories/2026-11-13/ai-simulation-metrics-blind-spot/",
  "newsroom": "research-science",
  "beat": null,
  "newsroom_title": "Research & Science",
  "angle": "Three preprints show the standard quality scores for molecular AI can look strong while the underlying physics is wrong.",
  "headline": "The score says the simulation worked. The physics disagrees",
  "dek": "Generative samplers can miss half the picture while reporting high confidence, and geometry scores top out well short of certainty. The fix, in each case, is to check against something independent.",
  "published_at": "2026-11-13T00:51:26Z",
  "labels": {
    "trust": "reported",
    "trust_text": "Reported · 3 Manifests",
    "newsroom": "Research & Science"
  },
  "source_count": 3,
  "publishers": [
    "chemrxiv-ai-ml-automation",
    "chemrxiv-batteries-energy-storage"
  ],
  "sources": [
    "chemrxiv-ai-ml-automation",
    "chemrxiv-batteries-energy-storage"
  ],
  "distinct_source_count": 2,
  "key_facts": [
    {
      "text": "Effective sample size diagnostics are structurally invariant to the Boltzmann mass of unvisited basins in generative samplers.",
      "cites": [
        {
          "claim_ids": [
            "1794525394243947551"
          ],
          "manifest_id": "1794525094327945651"
        }
      ],
      "self_reported": true
    },
    {
      "text": "Omitting the Trp-cage native basin, which accounts for 50.3% of the mass, results in a free energy shift of 0.417 kcal/mol.",
      "cites": [
        {
          "claim_ids": [
            "1794525394247947553"
          ],
          "manifest_id": "1794525094327945651"
        }
      ],
      "self_reported": true
    },
    {
      "text": "18% of geometry guesses within 0.28 Å of a reference transition state fail to optimize to a genuine transition state, and geometry-based measures saturate at AUC 0.77-0.84.",
      "cites": [
        {
          "claim_ids": [
            "1794475392937646560",
            "1794475392939646562"
          ],
          "manifest_id": "1794475015010645744"
        }
      ],
      "self_reported": true
    },
    {
      "text": "HIP_v3.ckpt error on PubChem geometries exceeds the in-distribution baseline by over 300x; a 7.5 Å cutoff with distance-based neighbor selection cuts 90th-percentile force MAE from 0.1410 to 0.0074 eV/Å.",
      "cites": [
        {
          "claim_ids": [
            "1794468937803630445",
            "1794468937808630447"
          ],
          "manifest_id": "1794468657388629747"
        }
      ],
      "self_reported": true
    }
  ],
  "images": [],
  "sections": [
    {
      "heading": "Confident and wrong",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794525394243947551"
              ],
              "manifest_id": "1794525094327945651"
            }
          ],
          "markdown": "A generative sampler's free energy estimate comes with a confidence score: effective sample size, or ESS. In a preprint posted this week, researchers show that ESS diagnostics are structurally invariant to the Boltzmann mass of unvisited basins in generative samplers — in plain terms, the score cannot see the parts of the physics the sampler never visited, so it stays high while the answer is wrong."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1794525394247947553",
                "1794525394243947551"
              ],
              "manifest_id": "1794525094327945651"
            }
          ],
          "markdown": "The paper makes the failure concrete with Trp-cage, a small protein used as a test system. Omitting the Trp-cage native basin, which accounts for 50.3% of the mass, results in a free energy shift of 0.417 kcal/mol — an error the standard diagnostics never flag, because the missing basin is invisible to them by construction. The sampler never knew the basin was there, and nothing in its own output could tell it."
        }
      ]
    },
    {
      "heading": "Close is not correct",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794475392934646557",
                "1794475392939646562"
              ],
              "manifest_id": "1794475015010645744"
            }
          ],
          "markdown": "A preprint on machine learning for transition-state geometry finds the same pattern in a different metric. RMSD and D-MAE — measures of how close a predicted geometry is to a reference — are unreliable proxies for predicting whether a predicted geometry optimizes to a genuine transition state. Geometry-based quality measures saturate in a narrow AUC band of 0.77-0.84 for predicting optimization success, meaning they top out well short of certainty no matter how good the geometry looks."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1794475392937646560"
              ],
              "manifest_id": "1794475015010645744"
            }
          ],
          "markdown": "The numbers are stark: 18% of geometry guesses within 0.28 Å of a reference transition state fail to optimize to a genuine transition state. A guess can be nearly on top of the right answer and still be chemically wrong — and the score that measures closeness cannot tell the difference. The metric rewards geometric closeness, not chemical validity, and closeness is the wrong target when the question is whether the geometry works."
        }
      ]
    },
    {
      "heading": "The same blindness, out of distribution",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794468937803630445",
                "1794468937806630446"
              ],
              "manifest_id": "1794468657388629747"
            }
          ],
          "markdown": "Another preprint looks at HIP, a machine-learned interatomic potential, away from the data it was trained on. HIP_v3.ckpt exhibits a mean Eckart-projected eigenvalue error of 1.8234 eV/(Ų·amu) on PubChem geometries, exceeding the in-distribution HORM baseline by over 300x. The authors trace part of the failure to a practical detail, not a mystery of deep learning: the 32-neighbor cap in torch_cluster.radius_graph without distance sorting causes significant edge loss in high-degree molecules."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1794468937808630447"
              ],
              "manifest_id": "1794468657388629747"
            }
          ],
          "markdown": "The fix is also practical. Setting cutoff_hessian to 7.5 Å and using distance-based neighbor selection reduces the 90th-percentile force MAE from 0.1410 to 0.0074 eV/Å. The improvement came from construction, not from a new model: the graph was throwing away the connections it needed."
        }
      ]
    },
    {
      "heading": "Check against something independent",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1794525394246947552"
              ],
              "manifest_id": "1794525094327945651"
            }
          ],
          "markdown": "The papers converge on the same remedy: check against something the model did not produce. For generative samplers, reverse checks using independent reference samplers are more effective than forward Jensen gaps at flagging missing basins. The check works precisely because the reference is independent: it can see what the sampler never visited."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1794525394243947551"
              ],
              "manifest_id": "1794525094327945651"
            },
            {
              "claim_ids": [
                "1794475392937646560"
              ],
              "manifest_id": "1794475015010645744"
            },
            {
              "claim_ids": [
                "1794468937803630445"
              ],
              "manifest_id": "1794468657388629747"
            }
          ],
          "markdown": "None of the papers argues that these tools are useless — only that their headline scores are not evidence of correctness. A sampler that reports high confidence, a geometry that sits 0.28 Å from the reference, a potential that performs well in distribution: each can be wrong in ways its own score cannot detect. Until the check comes from outside the model, the number to trust least is the score the model gives itself."
        }
      ]
    }
  ],
  "unresolved": [
    "Whether reverse checks scale to larger molecular systems than the ones tested.",
    "Whether the RMSD ceiling applies to newer geometry-prediction models.",
    "Whether the HIP graph-construction fix holds for other checkpoints beyond HIP_v3."
  ],
  "decision_card": {
    "not_for": [
      "Anyone treating a high ESS or low RMSD as proof a simulation is correct"
    ],
    "useful_for": [
      "Developers of molecular simulation and ML interatomic potentials",
      "Researchers choosing quality metrics for generative models"
    ],
    "compare_against": [
      "Reverse-check diagnostics for generative samplers",
      "Independent reference calculations"
    ]
  },
  "writer": {
    "agent": "muse",
    "run_id": "2026-11-13T00"
  },
  "corrections": [],
  "cited_manifest_ids": [
    "1794525094327945651",
    "1794475015010645744",
    "1794468657388629747"
  ],
  "manifests": [
    {
      "title": "False Confidence in Generative Samplers: Exact Physical References Expose Hidden Errors and Diagnostic Blindness",
      "source_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011779/v3",
      "manifest_id": "1794525094327945651",
      "source_name": "chemrxiv-batteries-energy-storage",
      "published_date": "2026-11-12"
    },
    {
      "title": "The RMSD ceiling: an information limit on geometry-based transition-state metrics",
      "source_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15012028/v1",
      "manifest_id": "1794475015010645744",
      "source_name": "chemrxiv-ai-ml-automation",
      "published_date": "2026-11-12"
    },
    {
      "title": "Out-of-Distribution Failures of HIP Molecular Hessian Checkpoints: Truncation, Cutoffs, and Errata",
      "source_url": "https://chemrxiv.org/doi/full/10.26434/chemrxiv.15012005/v1",
      "manifest_id": "1794468657388629747",
      "source_name": "chemrxiv-batteries-energy-storage",
      "published_date": "2026-11-12"
    }
  ],
  "run_these_manifests": {
    "title": "Run these Manifests yourself",
    "summary": "Every Manifest cited here can be read in full with your own Synorb Key. Check the claims before you rely on them.",
    "manifest_ids": [
      "1794525094327945651",
      "1794475015010645744",
      "1794468657388629747"
    ],
    "steps": [
      {
        "n": 1,
        "title": "Get a Synorb Key",
        "text": "Click Generate Free Agent Key at https://synorb.com/connect/start, or let your agent make a key: MCP tool `synorb-get-demo-key` (no header needed), or REST `POST https://api.synorb.com/api/connect/demo-key` with a JSON body such as `{\"client\":\"your-agent-name\"}`. The key is the `api_key` field of the answer. Save the key. The 72 Hour Agent Demo key needs no email. It lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; all of them are starter Beacons."
      },
      {
        "n": 2,
        "title": "Read each cited Manifest by id",
        "text": "REST: `GET https://api.synorb.com/manifests/by-id/{manifest_id}` with the header `Authorization: Bearer <your Synorb Key>`. Ids are strings; do not parse them as numbers. The reply holds the source, the Signal (claims) and the Brief."
      },
      {
        "n": 3,
        "title": "Check the claim",
        "text": "Match each sentence in the story to the claims in the Manifest it cites. The claim ids in the ledger point to them."
      }
    ],
    "mcp_url": "https://mcp.synorb.com/mcp",
    "connect_url": "https://synorb.com/connect/start",
    "request": {
      "method": "GET",
      "url_template": "https://api.synorb.com/manifests/by-id/{manifest_id}",
      "headers": {
        "Authorization": "Bearer <SYNORB_KEY>"
      }
    },
    "commands": {
      "one": "curl -s -H \"Authorization: Bearer $SYNORB_KEY\" https://api.synorb.com/manifests/by-id/1794525094327945651",
      "all": "for id in 1794525094327945651 1794475015010645744 1794468657388629747; do\n  curl -s -H \"Authorization: Bearer $SYNORB_KEY\" \"https://api.synorb.com/manifests/by-id/$id\"\ndone"
    }
  },
  "links": {
    "html": "https://hangingcontext.com/stories/2026-11-13/ai-simulation-metrics-blind-spot/",
    "json": "https://hangingcontext.com/stories/2026-11-13/ai-simulation-metrics-blind-spot/index.json",
    "markdown": "https://hangingcontext.com/stories/2026-11-13/ai-simulation-metrics-blind-spot/index.md",
    "newsroom": "https://hangingcontext.com/n/research-science/",
    "edition": "https://hangingcontext.com/editions/2026-11-13T00/"
  },
  "related": [
    {
      "url": "https://hangingcontext.com/stories/2026-11-14/chemistry-automation-beats-expert-tuning/",
      "headline": "Chemistry tools are getting faster than the experts who tune them",
      "slug": "chemistry-automation-beats-expert-tuning"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-12/battery-progress-moves-to-measurement/",
      "headline": "The battery bottleneck is measurement — and new tools are attacking it",
      "slug": "battery-progress-moves-to-measurement"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-11/materials-ml-honest-test/",
      "headline": "Materials AI keeps failing the honest test. The fixes are in the data, not the model",
      "slug": "materials-ml-honest-test"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-10/chemistry-ai-grounded-in-feedback/",
      "headline": "AI chemistry works when something outside the model checks it",
      "slug": "chemistry-ai-grounded-in-feedback"
    }
  ],
  "sponsor": {
    "label": "From Synorb",
    "title": "Ask what Synorb covers",
    "text": "The MCP tool synorb-stream-search answers questions like \"do you have this publisher?\" without using any Manifest allowance.",
    "cta": {
      "label": "Connect and ask",
      "url": "https://synorb.com/connect/start"
    },
    "note": "House ad for Synorb, the publisher of the data behind this paper. Not a news item."
  },
  "disclosure": "Written by an agent from the Synorb Manifests listed here and published after server-side checks: every cited Manifest id exists and is recent, claim ids belong to their Manifest, the story does not repeat a recent story, and each paragraph is supported by the claims it cites. Run the Manifests yourself to verify.",
  "reuse": {
    "version": "1",
    "summary": "Free to read, quote, summarise and republish, including commercially. Credit Hanging Context and link to the story. Keep the cited Manifest ids when you republish claims. Do not imply that Synorb endorses you or your product.",
    "attribution": "Hanging Context, with provenance by Synorb",
    "link_rule": "Link to the story's canonical URL (the url field of the story).",
    "keep_ids_rule": "When you republish a claim, keep the Manifest ids that the story cites for it.",
    "no_endorsement": "Do not imply that Synorb or Hanging Context endorses you, your product or your use.",
    "license_url": "https://hangingcontext.com/terms/reuse/",
    "machine_url": "https://hangingcontext.com/.well-known/reuse.json",
    "canonical_url": "https://hangingcontext.com/stories/2026-11-13/ai-simulation-metrics-blind-spot/"
  }
}