SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2607.00339 · 2026

TRACE: State-Aware Query Processing over Temporal Evidence Graphs for Conversational Data

Yu Wang, Chenbin Zhang, Hao Miao, Maolin Wang, Haoran Yang, Jiguang Shen, Zichun Liu, Baiyuan Qiu

arXiv · PDF · Open in the Atlas

Code that ran

12 of 14 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 2 violated it and 0 have no contract verdict; 9 failures of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2607.00339.json · badge: /api/ran/2607.00339.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
MorinWang/TRACE canonical (classifier label)
classifier basis: census
12 ran · 2 no recorded run
FunctionStatusWhere it lives
build_nickname_map Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/event_extractor.py
code served (permissive licence) · get_code("ae721796fa294a27")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re
from typing import Dict, List, Optional, Set, Tuple


class MockNote:
    """Stand-in class for a memory note with content attribute."""
    def __init__(self, content: str):
        self.content = content


# Create a memories dict with notes that contain "Speaker X says:" patterns
memories = {
    "note_1": MockNote("Speaker Alice says: hello there"),
    "note_2": MockNote("Speaker Bob says: goodbye"),
    "note_3": MockNote("Speaker Alexandra says: nice to meet you"),
    "note_4": MockNote("Some other content without speaker pattern"),
}

ARGS = {
    "memories": memories,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
extract_json_block Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/parsing_utils.py
code served (permissive licence) · get_code("018899f9ff80edc8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

# The function extract_json_block takes a text string parameter
# and extracts the outermost JSON object from it using find() and rfind()
# It returns a string

# Create test cases with various JSON structures
text_with_json = 'Some prefix text {"key": "value", "nested": {"inner": "data"}} some suffix'

ARGS = {
    'text': text_with_json
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
format_turn_for_memory Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/dataset_adapter.py
code served (permissive licence) · get_code("2a670410f656592c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from typing import Mapping

torch.manual_seed(0)

# Construct a simple turn dictionary that satisfies the Mapping[str, str] interface
turn = {
    'role': 'user',
    'content': 'Hello, how are you?'
}

# Define ARGS with the single parameter required by format_turn_for_memory
ARGS = {
    'turn': turn
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
inject_topic_nodes Ran
ran, honoured the contract we drafted
MorinWang/TRACE/trace/session_injector.py
code served (permissive licence) · get_code("c95a57dc8a89713b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import logging
from unittest.mock import Mock, MagicMock
import networkx as nx

logging.basicConfig(level=logging.INFO)

# Create a mock CausalGraph class
class MockCausalGraph:
    def __init__(self):
        self.graph = nx.DiGraph()
        self._sessions = {
            "session_1": "User logged in and browsed products",
            "session_2": "User made a purchase transaction"
        }
        self._topics = {}
    
    def add_topic(self, topic_dict):
        topic_id = topic_dict.get('id', f'topic_{len(self._topics)}')
        self._topics[topic_id] = topic_dict
        self.graph.add_node(topic_id, **topic_dict)

# Create a mock Topic class
class MockTopic:
    def __init__(self, topic_id, name, sessions):
        self.id = topic_id
        self.name = name
        self.sessions = sessions
    
    def to_dict(self):
        return {
            'id': self.id,
            'name': self.name,
            'sessions': self.sessions
        }

# Create a mock LLM controller
class MockLLM:
    def get_completion(self, prompt):
        return "Topic clustering result"

# Mock the cluster_sessions_llm function
def mock_cluster_sessions_llm(session_summaries, llm, sample_idx):
    topics = [
        MockTopic('topic_0', 'Authentication', ['session_1']),
        MockTopic('topic_1', 'Commerce', ['session_2'])
    ]
    return topics

# Monkey patch the import
import sys
from types import ModuleType

trace_module = ModuleType('trace')
topic_clusterer = ModuleType('topic_clusterer')
topic_clusterer.cluster_sessions_llm = mock_cluster_sessions_llm
trace_module.topic_clusterer = topic_clusterer
sys.modules['trace'] = trace_module
sys.modules['trace.topic_clusterer'] = topic_clusterer

# Create arguments
graph = MockCausalGraph()
llm = MockLLM()
sample_idx = 0

ARGS = {
    'graph': graph,
    'llm': llm,
    'sample_idx': sample_idx
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
networkx — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize_lme_role Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/dataset_adapter.py
code served (permissive licence) · get_code("8993fe2601f2da08")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

ARGS = {
    'role': 'assistant'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize_participants Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/event_extractor.py
code served (permissive licence) · get_code("675f4e3a054bcf7f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
from typing import Dict, List, Optional, Set, Tuple

torch.manual_seed(0)

# Construct valid arguments for normalize_participants
participants = ["alice", "Bob", "CHARLIE", "alice", "bob"]
nickname_map = {
    "alice": "Alice Smith",
    "bob": "Robert Johnson",
    "charlie": "Charles Brown"
}

ARGS = {
    "participants": participants,
    "nickname_map": nickname_map
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_longmemeval_date Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/dataset_adapter.py
code served (permissive licence) · get_code("2765833e486981a5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned datetime. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re
from datetime import datetime

def parse_longmemeval_date(value: str):
    """Return a sortable datetime for LongMemEval timestamps."""
    if not value:
        return datetime.min
    cleaned = re.sub('\\s+\\([A-Za-z]{3}\\)', '', value.strip())
    for fmt in ('%Y/%m/%d %H:%M', '%Y/%m/%d', '%Y-%m-%d %H:%M', '%Y-%m-%d'):
        try:
            return datetime.strptime(cleaned, fmt)
        except ValueError:
            continue
    return datetime.min

# Test cases for different date formats
ARGS = {
    "value": "2023/12/25 14:30"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type datetime — a datetime output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a datetime output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
strip_markdown_fences Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/llm_text_io.py
code served (permissive licence) · get_code("21854723dac05890")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def strip_markdown_fences(text: str) -> str:
    """Remove ```json ... ``` or ``` ... ``` fences from LLM output."""
    text = text.strip()
    text = re.sub('^```(?:json)?\\s*\\n?', '', text, flags=re.MULTILINE)
    text = re.sub('\\n?\\s*```$', '', text, flags=re.MULTILINE)
    return text.strip()

# Setup test arguments
ARGS = {
    "text": "```json\n{\"key\": \"value\"}\n```"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
strip_markdown_fences Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/parsing_utils.py
code served (permissive licence) · get_code("5440cdf99219bc9c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import re

def strip_markdown_fences(text: str) -> str:
    """Remove ```json ... ``` fences from LLM output."""
    text = text.strip()
    text = re.sub('^```(?:json)?\\s*\\n?', '', text, flags=re.MULTILINE)
    text = re.sub('\\n?```\\s*$', '', text, flags=re.MULTILINE)
    return text.strip()

# Test cases with various markdown fence scenarios
ARGS = {
    "text": "```json\n{\"key\": \"value\"}\n```"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
try_parse_date Ran
ran; our contract draft was wrong, not the code
MorinWang/TRACE/trace/parsing_utils.py
code served (permissive licence) · get_code("70605eacf8e400d2")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned datetime. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from datetime import datetime
from typing import Optional

def try_parse_date(s: str) -> Optional[datetime]:
    """Try to parse a string as an absolute date."""
    for fmt in ('%Y-%m-%d', '%Y%m%d', '%Y/%m/%d', '%B %d, %Y', '%b %d, %Y', '%Y-%m-%dT%H:%M', '%Y%m%d%H%M'):
        try:
            return datetime.strptime(s.strip(), fmt)
        except ValueError:
            continue
    return None

ARGS = {
    's': '2024-01-15'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type datetime — a datetime output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a datetime output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
validate_edge_type Ran
ran, violated the contract we drafted
MorinWang/TRACE/trace/event_schema.py
code served (permissive licence) · get_code("860897df3ea15ce4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

# The function validate_edge_type takes a single parameter: edge_type (a string)
# It checks if the edge_type is in the VALID_EDGE_TYPES set
# We need to provide a valid edge_type string

ARGS = {
    'edge_type': 'causes'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
validate_event_type Ran
ran, violated the contract we drafted
MorinWang/TRACE/trace/event_schema.py
code served (permissive licence) · get_code("0cad76b60328f8a8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

ARGS = {
    "event_type": "action"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_analyze_content Not yet run
unverified (harvested, no recorded run)
MorinWang/TRACE/trace/llm_text_io.py
code served (permissive licence) · get_code("02583dd626081175")
How it ran

no invocation recorded.

parse_with_json_fallback Not yet run
unverified (harvested, no recorded run)
MorinWang/TRACE/trace/llm_text_io.py
code served (permissive licence) · get_code("fd3c19ff87593021")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 12 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
parse_analyze_content 2 other papers
2 dated earlier
2502.12110 (2025-02)
A-Mem: Agentic Memory for LLM Agents
2606.07402 (2026-06)
M 3 Exam: Benchmarking Multimodal Memory for Realistic User-Agent Interactions
strip_markdown_fences 2 other papers
2 dated earlier
2502.12110 (2025-02)
A-Mem: Agentic Memory for LLM Agents
2606.07402 (2026-06)
M 3 Exam: Benchmarking Multimodal Memory for Realistic User-Agent Interactions

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Conversational data is increasingly used as a persistent source of user state for long-running assistants and AI agents. However, querying this data remains challenging because conversations naturally evolve: plans are revised, preferences change, and later messages frequently supersede or contradict earlier information. Existing long-memory pipelines largely treat memories as independent text or vector objects. This approach often retrieves semantically similar but stale evidence, offering limited support for state-aware reasoning. To address this problem, we present TRACE, a query processing framework over temporal evidence graphs for evolving conversational data. TRACE models conversations as a hierarchical graph spanning events, sessions, and topics, enriched with typed temporal, causal, update, and contradiction relations. Crucially, the framework maintains validity annotations so obsolete facts remain accessible for historical queries but are discounted for current-state answers. At query time, TRACE combines vector-based note retrieval with graph-guided evidence search, generating validity-aware support paths and a hybrid context for answer generation. This design separates lexical recall from evidence reconstruction, enabling bounded query-time reasoning over long conversational histories. Experiments on long-conversation query-answering (QA) benchmarks show that TRACE improves temporal and multi-hop reasoning, with ablations highlighting the importance of hierarchy, update-aware seeding, and path-grounded evidence. Source code is provided at https://github.com/MorinWang/TRACE.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2607.00339")
get_code_for_paper("2607.00339")
have("2607.00339")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2607.00339.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2607.00339.svg)](https://syntology.ai/paper/2607.00339)

Connect an agent — have() is free.