Pith Number

pith:MHY4IJOM

pith:2026:MHY4IJOM5AWQZDUXMEKYT2CVDX

not attested not anchored not stored refs resolved

Reinforcement Learning for Tool-Calling Agents in Fast Healthcare Interoperability Resources (FHIR)

Jan P. Bremer, Marius S. Knorr, Nils Schweingruber, Robert M\"uller

Reinforcement learning post-training raises a Qwen3-8B model to 77 percent correctness on FHIR clinical queries, surpassing larger closed models.

arxiv:2605.14126 v1 · 2026-05-13 · cs.LG · cs.AI

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{MHY4IJOM5AWQZDUXMEKYT2CVDX}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

Empirically, our approach improves answer correctness from 50% (o4-mini) to 77% on FHIR-AgentBench using a smaller and cheaper Qwen3-8B model.

C2weakest assumption

That an LLM Judge can supply reliable execution-grounded rewards without bias or systematic errors in judging multi-step FHIR traversals.

C3one line summary

RL post-training lifts answer correctness on FHIR-AgentBench from 50% (o4-mini) to 77% with a cheaper Qwen3-8B CodeAct agent.

References

66 extracted · 66 resolved · 8 Pith anchors

[1] Journal of the American Medical Informatics Association , author = 2017 · doi:10.1093/jamia/ocx080

[2] Gut and Liver , author = 2024 · doi:10.5009/gnl230272

[3] AMA journal of ethics , author = 2017 · doi:10.1001/journalofethics.2017.19.3.stas1-1703

[4] WIREs Computational Statistics , author = 2021 · doi:10.1002/wics.1549

[5] iScience , author = 2024 · doi:10.1016/j.isci.2024.109713

Receipt and verification

First computed	2026-05-17T23:39:11.844744Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

61f1c425cce82d0c8e97611589e8551de182451e0ffd6adfbd7199d79f0f8a12

Aliases

arxiv: 2605.14126 · arxiv_version: 2605.14126v1 · doi: 10.48550/arxiv.2605.14126 · pith_short_12: MHY4IJOM5AWQ · pith_short_16: MHY4IJOM5AWQZDUX · pith_short_8: MHY4IJOM

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/MHY4IJOM5AWQZDUXMEKYT2CVDX \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 61f1c425cce82d0c8e97611589e8551de182451e0ffd6adfbd7199d79f0f8a12

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "52e8853a78ec8c87734ac98a6b25bb71dc31ecc07f94480c1c465ac50bb3f52c",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-13T21:27:21Z",
    "title_canon_sha256": "567d44002f2e6213549591e3fea47891316a6c7be41cba3fa01d3f383819bb43"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.14126",
    "kind": "arxiv",
    "version": 1
  }
}