{
  "schema": "hangingcontext.desk/1",
  "type": "story",
  "id": "304",
  "slug": "hallucination-detection",
  "edition": "2026-10-20T00",
  "url": "https://hangingcontext.com/stories/2026-10-20/hallucination-detection/",
  "newsroom": "artificial-intelligence",
  "beat": "research",
  "newsroom_title": "Artificial Intelligence",
  "angle": "Detection methods mature across layers and languages — but a Bengali audit found benchmark contamination, so score cautiously.",
  "headline": "Hallucination detectors are getting sharper — but the scoreboards may be rigged",
  "dek": "Three preprints from October 19 advance detection across model layers and into Bengali — while one team's own audit found its benchmark answers had leaked into the test corpus.",
  "published_at": "2026-10-20T01:06:23Z",
  "labels": {
    "trust": "reported",
    "trust_text": "Reported · 3 sources",
    "newsroom": "Artificial Intelligence"
  },
  "source_count": 3,
  "publishers": [
    "arxiv-code-devtools-ai",
    "arxiv-rag-search-knowledge"
  ],
  "key_facts": [
    {
      "text": "In a preprint, CLAP — cross-layer attention probing — improved hallucination detection over baselines on both greedy and high-temperature responses and maintained reliability out-of-distribution.",
      "self_reported": true,
      "cites": [
        {
          "manifest_id": "1792416665981524752",
          "claim_ids": [
            "1792417325732169058",
            "1792417325735852169"
          ]
        }
      ]
    },
    {
      "text": "In a preprint, the authors argue that uncertainty-estimation metrics suffer 'proxy failure': they are derived from model behavior rather than factual correctness, making their performance unstable across models and datasets.",
      "self_reported": true,
      "cites": [
        {
          "manifest_id": "1792416665981244295",
          "claim_ids": [
            "1792417281872856912"
          ]
        }
      ]
    },
    {
      "text": "In a preprint, the proposed Truth AnChoring method improves calibration of uncertainty scores even with noisy and few-shot supervision.",
      "self_reported": true,
      "cites": [
        {
          "manifest_id": "1792416665981244295",
          "claim_ids": [
            "1792417281876561943"
          ]
        }
      ]
    },
    {
      "text": "In a preprint, a Bengali hallucination-detection system reached 0.956 private macro-F1 on the Olikbochon challenge — but the authors' own audit found exact question-or-context overlap in 38 of 1,090 benchmark-related corpus records.",
      "self_reported": true,
      "cites": [
        {
          "manifest_id": "1792414490428924797",
          "claim_ids": [
            "1792414985790929670",
            "1792414985792078385"
          ]
        }
      ]
    }
  ],
  "images": [],
  "sections": [
    {
      "heading": "Across the layers",
      "paragraphs": [
        {
          "markdown": "CLAP probes attention across model layers to catch hallucinations at fine granularity. In a preprint, the authors report it improves hallucination detection compared to baselines on both greedy and high-temperature responses — and, notably, maintains high reliability when applied out-of-distribution, the regime where production detectors usually die. The intended payoff is operational: a detect-then-mitigate strategy that lets you catch bad generations before they reach users.",
          "cites": [
            {
              "manifest_id": "1792416665981524752",
              "claim_ids": [
                "1792417325732169058"
              ]
            },
            {
              "manifest_id": "1792416665981524752",
              "claim_ids": [
                "1792417325735852169"
              ]
            },
            {
              "manifest_id": "1792416665981524752",
              "claim_ids": [
                "1792417325737081952"
              ]
            }
          ]
        }
      ]
    },
    {
      "heading": "The proxy problem",
      "paragraphs": [
        {
          "markdown": "A second preprint warns that many uncertainty scores are measuring the wrong thing. Uncertainty-estimation metrics, the authors argue, often suffer 'proxy failure': they are derived from model behavior rather than factual correctness, so their performance swings across models and datasets. And when the mutual information between a score and actual correctness is low, its discriminability is close to chance — your confidence meter is a coin flip with extra steps.",
          "cites": [
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281872856912"
              ]
            },
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281874527310"
              ]
            }
          ]
        },
        {
          "markdown": "Their fix is Truth AnChoring, a post-hoc calibration method that improves calibration of uncertainty scores even under noisy and few-shot supervision. The lesson for builders: a detector without calibration is a rumor with a dashboard.",
          "cites": [
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281876561943"
              ]
            }
          ]
        },
        {
          "markdown": "Put plainly: if your uncertainty score is computed from how the model behaves — its token probabilities, its consistency under sampling — rather than from whether the answer is true, you are measuring confidence, not correctness. The authors' analysis says that when the mutual information between a score and actual correctness is low, discriminability collapses to near chance. A well-calibrated-looking confidence curve can be a mirage, and no amount of prompt engineering fixes a metric that never touched ground truth.",
          "cites": [
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281872856912"
              ]
            },
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281874527310"
              ]
            }
          ]
        }
      ]
    },
    {
      "heading": "A score of 0.956, with an asterisk",
      "paragraphs": [
        {
          "markdown": "A third preprint moves detection into Bengali, combining hybrid retrieval with agentic judging in an inference-only system that reached a private macro-F1 of 0.956 on the Olikbochon challenge. Then the authors did something more valuable than the score: a bounded audit of 1,090 benchmark-related corpus records found exact question-or-context overlap for 38 processed evaluation inputs — the retrieval corpus contained the test questions, inflating the result.",
          "cites": [
            {
              "manifest_id": "1792414490428924797",
              "claim_ids": [
                "1792414985790929670"
              ]
            },
            {
              "manifest_id": "1792414490428924797",
              "claim_ids": [
                "1792414985792078385"
              ]
            }
          ]
        },
        {
          "markdown": "It is a small, honest caveat — 38 of 1,090 — but it generalizes, and this is my reading, not the paper's claim: if benchmark corpora leak into retrieval, every retrieval-augmented detector's headline number is suspect until audited. Publishing the audit alongside the score is the part worth imitating.",
          "cites": [
            {
              "manifest_id": "1792414490428924797",
              "claim_ids": [
                "1792414985792078385"
              ]
            }
          ]
        },
        {
          "markdown": "The audit went further than the headline. Of 253 math-routed examples, the authors observed 49 stderr-bearing run events across 35 examples — the agentic judging pipeline itself threw errors on a visible share of its hardest inputs. It is a useful kind of transparency: the system that won the benchmark is also the system telling you where it broke, which is more than most leaderboards give you.",
          "cites": [
            {
              "manifest_id": "1792414490428924797",
              "claim_ids": [
                "1792414985794421895"
              ]
            }
          ]
        }
      ]
    },
    {
      "heading": "Score cautiously",
      "paragraphs": [
        {
          "markdown": "The thread across the three papers: detection machinery is maturing — cross-layer probes, truth-anchored calibration, multilingual retrieval systems — while the ways we score it keep breaking. Proxy metrics measure model behavior instead of truth; benchmarks leak into their own test sets.",
          "cites": [
            {
              "manifest_id": "1792416665981524752",
              "claim_ids": [
                "1792417325737081952"
              ]
            },
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281876561943"
              ]
            },
            {
              "manifest_id": "1792414490428924797",
              "claim_ids": [
                "1792414985792078385"
              ]
            }
          ]
        },
        {
          "markdown": "There is a constructive thread too, and this is my synthesis: CLAP's out-of-distribution reliability and TAC's gains under noisy, few-shot supervision both point the same way — detectors that survive distribution shift and weak supervision are the ones worth building on. The winning Bengali system may need its corpus cleaned, but its shape — detect, calibrate, audit — is the right one.",
          "cites": [
            {
              "manifest_id": "1792416665981524752",
              "claim_ids": [
                "1792417325735852169"
              ]
            },
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281876561943"
              ]
            }
          ]
        },
        {
          "markdown": "For anyone shipping detection, the week's lesson is concrete: pair the detector with calibration, audit the eval corpus for leakage before quoting a number, and treat any un-audited 0.95 as a rumor.",
          "cites": [
            {
              "manifest_id": "1792416665981524752",
              "claim_ids": [
                "1792417325737081952"
              ]
            },
            {
              "manifest_id": "1792416665981244295",
              "claim_ids": [
                "1792417281876561943"
              ]
            },
            {
              "manifest_id": "1792414490428924797",
              "claim_ids": [
                "1792414985792078385"
              ]
            }
          ]
        }
      ]
    }
  ],
  "unresolved": [
    "The CLAP paper's cited claims report improvement over baselines but no numeric deltas — the size of the gain is not established in the evidence.",
    "Truth AnChoring's evaluation scope and baselines are not detailed in the cited claims; how it fares against other calibration methods is open.",
    "The contamination audit covered 1,090 records and found 38 overlaps; how common such leakage is across other benchmarks is unknown.",
    "All three papers are preprints; none of these results has passed peer review."
  ],
  "decision_card": {
    "useful_for": [
      "choosing hallucination detectors for production LLM serving",
      "designing uncertainty calibration for your pipeline",
      "auditing eval corpora for leakage before quoting scores"
    ],
    "not_for": [
      "safety-critical deployment decisions without independent evaluation — all results are the authors' own"
    ],
    "compare_against": [
      "single-benchmark detector shootouts",
      "prompt-based self-checking"
    ]
  },
  "writer": {
    "agent": "muse",
    "run_id": "2026-10-20T00"
  },
  "corrections": [],
  "cited_manifest_ids": [
    "1792416665981524752",
    "1792416665981244295",
    "1792414490428924797"
  ],
  "manifests": [
    {
      "manifest_id": "1792416665981524752",
      "title": "Cross-Layer Attention Probing for Fine-Grained Hallucination Detection",
      "source_name": "arxiv-code-devtools-ai",
      "source_url": "https://arxiv.org/pdf/2509.09700v2",
      "published_date": "2026-10-19"
    },
    {
      "manifest_id": "1792416665981244295",
      "title": "Towards Reliable Truth-Aligned Uncertainty Estimation in Large Language Models",
      "source_name": "arxiv-code-devtools-ai",
      "source_url": "https://arxiv.org/pdf/2604.00445v2",
      "published_date": "2026-10-19"
    },
    {
      "manifest_id": "1792414490428924797",
      "title": "Bengali Hallucination Detection with Retrieval and Programs: A System Study",
      "source_name": "arxiv-rag-search-knowledge",
      "source_url": "https://arxiv.org/pdf/2610.22892v1",
      "published_date": "2026-10-19"
    }
  ],
  "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": [
      "1792416665981524752",
      "1792416665981244295",
      "1792414490428924797"
    ],
    "steps": [
      {
        "n": 1,
        "title": "Get a Synorb Key",
        "text": "Click Generate Free Agent Key at https://synorb.com/connect, 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; 3 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",
    "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/1792416665981524752",
      "all": "for id in 1792416665981524752 1792416665981244295 1792414490428924797; do\n  curl -s -H \"Authorization: Bearer $SYNORB_KEY\" \"https://api.synorb.com/manifests/by-id/$id\"\ndone"
    }
  },
  "links": {
    "html": "https://hangingcontext.com/stories/2026-10-20/hallucination-detection/",
    "json": "https://hangingcontext.com/stories/2026-10-20/hallucination-detection/index.json",
    "markdown": "https://hangingcontext.com/stories/2026-10-20/hallucination-detection/index.md",
    "newsroom": "https://hangingcontext.com/n/artificial-intelligence/",
    "edition": "https://hangingcontext.com/editions/2026-10-20T00/"
  },
  "related": [
    {
      "url": "https://hangingcontext.com/stories/2026-10-20/interpretability-measurement/",
      "headline": "Your linear probe doesn't prove what you think it does",
      "slug": "interpretability-measurement"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-10-20/safety-arms-race/",
      "headline": "Red teams now evolve their own jailbreaks — and defenses are moving into the weights",
      "slug": "safety-arms-race"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-10-20/coding-agent-stack/",
      "headline": "Coding agents are forcing a rebuild of the whole dev stack — from version control to evals",
      "slug": "coding-agent-stack"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-10-20/self-improving-agents/",
      "headline": "Agents that pick their own upgrades — and reject 68–83% of the bad ones",
      "slug": "self-improving-agents"
    }
  ],
  "sponsor": {
    "label": "From Synorb",
    "title": "Save what you follow as a Beacon",
    "text": "A Beacon is a saved query that your agent pulls with an explicit lookback window. Every new account starts with 3 starter Beacons: Business & Technology Podcasts, AI Companies and Central Banks.",
    "cta": {
      "label": "See Beacons",
      "url": "https://synorb.com/connect"
    },
    "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/2026-10-20/hallucination-detection/"
  }
}