Publish reviewed Terminal Pro research aggregates

README.md

@@ -1,3 +1,82 @@
 ---
+pretty_name: DX Terminal Pro Research Aggregates
 license: cc-by-4.0
+language:
+- en
+tags:
+- tabular
+- agent-evaluation
+- research-aggregates
+- arxiv:2604.26091
+- arxiv:2609.05663
+configs:
+- config_name: market_behavior
+  default: true
+  data_files:
+  - split: observations
+    path: data/market_behavior.csv
+- config_name: mandate_compilation
+  data_files:
+  - split: observations
+    path: data/mandate_compilation.csv
 ---
+
+# DX Terminal Pro research aggregates
+
+Nine published research records from DX Research Group's Terminal Pro work. These are small aggregate evidence tables. They contain no participant-level decision logs or training trajectories.
+
+The two configurations have different units of analysis:
+
+| Configuration | Records | What a row represents |
+|---|---:|---|
+| `market_behavior` | 4 | A reported event or token-window aggregate from the bounded 21-day real-capital Terminal Pro deployment |
+| `mandate_compilation` | 5 | A reported pre-launch comparison or historical live setting gradient |
+
+The market-behavior rows describe the historical Base deployment. The mandate table contains three controlled pre-launch examples and two historical live gradients. Do not pool these as independent trials from one experiment. The earlier 2025 DX Terminal simulation is not included.
+
+## Load the tables
+
+```python
+from datasets import load_dataset
+
+market = load_dataset(
+    "DXRG/dx-terminal-pro-research-aggregates",
+    "market_behavior",
+    split="observations",
+)
+controls = load_dataset(
+    "DXRG/dx-terminal-pro-research-aggregates",
+    "mandate_compilation",
+    split="observations",
+)
+```
+
+
+## Sources and provenance
+
+Primary paper: [Operating-Layer Controls for Onchain Language-Model Agents Under Real Capital, arXiv:2604.26091v1](https://arxiv.org/abs/2604.26091v1). The [subsequent two-fleet paper](https://arxiv.org/abs/2609.05663v1) supplies the broader research context; these two configurations are Terminal Pro material.
+
+The market CSV preserves the four records from [DXRG's published observations](https://www.dxrg.ai/research/dx-terminal-pro-market-behavior-observations.v1.csv), version 1.0.0. The control CSV preserves all five records and original columns from [the mandate-compilation evidence table](https://www.dxrg.ai/research/trading-agent-mandate-compilation-evidence.v1.csv), version 1.0.0, and adds `measure_scope`, `comparison_sample_size` and `sample_size_note` to clarify denominators. The original metadata objects are preserved in `source_context/`.
+
+This release republishes reported measurements. It does not rerun the underlying analyses or independently validate the paper's findings. Dataset release version: 1.0.0.
+
+## Reading the measurements
+
+- The fabricated-rule comparison measures sell decisions in the affected pre-launch populations. Its before/after percentages describe a combined intervention, not an isolated memory-label effect.
+- The paper's aggregate diagnosis of 4,900 reasoning traces is not the denominator for every comparison. Exact per-arm counts are not supplied for these records; missing sample-size values remain null.
+- Trading Activity is a percentage of invocations that produce trade actions. Trade Size is a share of available ETH used per trade. Their similar percentage notation does not make them the same measure.
+- The two-sided-flow percentage is trade-weighted and uses five-minute windows for the same token. A two-sided window contains at least one buy and one sell.
+- A sell cascade requires at least ten vaults selling the same token within ten minutes. The largest reported cascade and the total cascade count are different aggregates.
+- Missing numeric cells mean a field is not applicable or is not supplied in that source record. They must not be converted to zero.
+
+The source locators, definitions and evidence-boundary columns accompany the values. No new causal effect, statistical significance, prevalence estimate or performance result is inferred by this packaging.
+
+## Scope and exclusions
+
+No wallet, vault, user, account, agent, order or trade identifiers are included. User strategies, chat, model prompts and reasoning, per-participant returns and individual transaction records are excluded. The separate internal EVM construction evaluation and internal model-selection screen are also excluded.
+
+The data describes a bounded historical experiment. It is not evidence of current DXAP Alpha performance, a general trading advantage, safety, or transfer to other venues. The published record includes limitations and negative findings; readers should use the linked papers for their full context.
+
+## License and citation
+
+License: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/), consistent with the associated papers. Attribute DX Research Group and the source paper when reusing the tables. Cite the original frozen paper version for a reported result and this dataset version for the packaged tables.

data/mandate_compilation.csv

@@ -0,0 +1,6 @@
+test_id,evaluation_design,intervention_or_setting,fixed_components,observed_measure,result,interpretation,boundary,source_location,measure_scope,comparison_sample_size,sample_size_note
+fee-reading-order,controlled-prelaunch-test,Move the unchanged 2.3% fee sentence from paragraph eight to paragraph one,Model; wording; market data; evaluation setting,Fee citation in sampled reasoning traces,3% to 74%,Rendering order changed use of a fixed fact in the evaluated traces,Reasoning-trace behavior; not a return or execution result,Control-Loop Results,Sampled reasoning traces in the fee-position comparison,,Not supplied for this comparison in the linked public source; the aggregate 4900-trace diagnosis is not a per-arm denominator
+fabricated-sell-rule-compound-intervention,controlled-prelaunch-test,Label prior reasoning as context and prohibit invented named rules,Affected controlled test population and stated harness evaluation,Traces containing fabricated sell rules,57% to 3%,The compound harness intervention reduced fabricated policy in the affected traces,Compound intervention; attribution cannot be assigned to memory labeling alone,Control-Loop Results and Table 4,Sell decisions in the affected pre-launch populations,,Not supplied for this comparison in the linked public source; the aggregate 4900-trace diagnosis is not a per-arm denominator
+structured-tokenomics-context,controlled-prelaunch-test,Lead with the reap payout and provide structured tokenomics context,Affected controlled pre-launch population and stated evaluation protocol,Capital deployment,42.9% to 78.0%,Representation and ordering changed capital deployment in the affected population,Deployment behavior; not profitability,Control-Loop Results and Table 4,Capital deployed in the affected pre-launch populations,,Not supplied for this comparison in the linked public source; the aggregate 4900-trace diagnosis is not a per-arm denominator
+trading-activity-setting-gradient,bounded-historical-live-gradient,Trading Activity structured setting,Frozen live model and runtime across the bounded deployment,Share of invocations producing trade actions,2.8% to 16.8% across settings,The structured setting mapped to a sixfold frequency spread in the deployment,Observed live gradient; not a randomized ablation or evidence that higher activity was better,Production Behavior,Share of invocations producing trade actions by setting,,Not supplied for this comparison in the linked public source; the aggregate 4900-trace diagnosis is not a per-arm denominator
+trade-size-setting-gradient,bounded-historical-live-gradient,Trade Size structured setting,Frozen live model and runtime across the bounded deployment,Share of available ETH used per trade,About 2% to about 95% across settings,The structured setting mapped to a large sizing gradient in the deployment,Observed live gradient; not a randomized ablation or evidence that larger trades were better,Production Behavior,Share of available ETH used per trade by setting,,Not supplied for this comparison in the linked public source; the aggregate 4900-trace diagnosis is not a per-arm denominator

data/market_behavior.csv

@@ -0,0 +1,5 @@
+observation_id,measurement,evidence_class,event_day,token_label,window_minutes,buying_vaults,active_vaults,sell_count,median_inter_agent_gap_seconds,cascade_count,minimum_selling_vaults,trade_share_percent,minimum_buys,minimum_sells,same_token_required,communication_state,source_locator,interpretation,evidence_boundary
+feet-one-hour-buy-cascade,buy-cascade,historical-live-deployment,3,FEET,60,1544,3454,,,,,,,,,"The vaults did not communicate with each other.","Production Behavior Under a Frozen Harness, page 9","Shared market state was sufficient for a concentrated buy event in the reported deployment.","One bounded 21-day deployment; no single-cause, prevalence, alpha, profitability, universal-safety, or transfer claim."
+poopcoin-largest-sell-cascade,largest-reported-sell-cascade,historical-live-deployment,,POOPCOIN,,,,438,9.5,,,,,,,,"Production Behavior Under a Frozen Harness, page 10","The row describes the largest sell cascade reported for this token in the deployment.","One bounded 21-day deployment; no single-cause, prevalence, alpha, profitability, universal-safety, or transfer claim."
+tournament-sell-cascade-count,sell-cascade-count,historical-live-deployment,,,10,,,,,3878,10,,,,true,,"Production Behavior Under a Frozen Harness, page 10","The count uses the paper's registered sell-cascade definition across the tournament.","One bounded 21-day deployment; no single-cause, prevalence, alpha, profitability, universal-safety, or transfer claim."
+two-sided-token-window-share,trade-share-in-two-sided-token-windows,historical-live-deployment,,,5,,,,,,,92.9,1,1,true,,"Production Behavior Under a Frozen Harness, page 10","The shared model produced both buy and sell flow when mandates, positions, and settings differed.","One bounded 21-day deployment; no single-cause, prevalence, alpha, profitability, universal-safety, or transfer claim."

source_context/mandate_compilation.json

@@ -0,0 +1,70 @@
+{
+  "schemaVersion": "dxrg-trading-agent-mandate-compilation-evidence/v1",
+  "version": "1.0.0",
+  "title": "DXRG Trading-Agent Mandate Compilation Evidence Table",
+  "publishedDate": "2026-07-21",
+  "publisher": "DX Research Group",
+  "canonicalPage": "https://www.dxrg.ai/blogs/trading-agent-mandate-compiler",
+  "methodology": "Each record states the evaluation design, intervention or setting, fixed components, observed measure, interpretation, and evidence boundary. Controlled pre-launch interventions and historical live setting gradients remain separate evidence classes.",
+  "source": {
+    "title": "Operating-Layer Controls for Onchain Language-Model Agents Under Real Capital",
+    "url": "https://arxiv.org/abs/2604.26091"
+  },
+  "records": [
+    {
+      "testId": "fee-reading-order",
+      "evaluationDesign": "controlled-prelaunch-test",
+      "interventionOrSetting": "Move the unchanged 2.3% fee sentence from paragraph eight to paragraph one",
+      "fixedComponents": "Model, wording, market data, and evaluation setting",
+      "observedMeasure": "Fee citation in sampled reasoning traces",
+      "result": "3% to 74%",
+      "interpretation": "Rendering order changed use of a fixed fact in the evaluated traces",
+      "boundary": "Reasoning-trace behavior, not a return or execution result",
+      "sourceLocation": "Control-Loop Results"
+    },
+    {
+      "testId": "fabricated-sell-rule-compound-intervention",
+      "evaluationDesign": "controlled-prelaunch-test",
+      "interventionOrSetting": "Label prior reasoning as context and prohibit invented named rules",
+      "fixedComponents": "Affected controlled test population and stated harness evaluation",
+      "observedMeasure": "Traces containing fabricated sell rules",
+      "result": "57% to 3%",
+      "interpretation": "The compound harness intervention reduced fabricated policy in the affected traces",
+      "boundary": "Compound intervention; attribution cannot be assigned to memory labeling alone",
+      "sourceLocation": "Control-Loop Results and Table 4"
+    },
+    {
+      "testId": "structured-tokenomics-context",
+      "evaluationDesign": "controlled-prelaunch-test",
+      "interventionOrSetting": "Lead with the reap payout and provide structured tokenomics context",
+      "fixedComponents": "Affected controlled pre-launch population and stated evaluation protocol",
+      "observedMeasure": "Capital deployment",
+      "result": "42.9% to 78.0%",
+      "interpretation": "Representation and ordering changed capital deployment in the affected population",
+      "boundary": "Deployment behavior, not profitability",
+      "sourceLocation": "Control-Loop Results and Table 4"
+    },
+    {
+      "testId": "trading-activity-setting-gradient",
+      "evaluationDesign": "bounded-historical-live-gradient",
+      "interventionOrSetting": "Trading Activity structured setting",
+      "fixedComponents": "Frozen live model and runtime across the bounded deployment",
+      "observedMeasure": "Share of invocations producing trade actions",
+      "result": "2.8% to 16.8% across settings",
+      "interpretation": "The structured setting mapped to a sixfold frequency spread in the deployment",
+      "boundary": "Observed live gradient, not a randomized ablation or evidence that higher activity was better",
+      "sourceLocation": "Production Behavior"
+    },
+    {
+      "testId": "trade-size-setting-gradient",
+      "evaluationDesign": "bounded-historical-live-gradient",
+      "interventionOrSetting": "Trade Size structured setting",
+      "fixedComponents": "Frozen live model and runtime across the bounded deployment",
+      "observedMeasure": "Share of available ETH used per trade",
+      "result": "About 2% to about 95% across settings",
+      "interpretation": "The structured setting mapped to a large sizing gradient in the deployment",
+      "boundary": "Observed live gradient, not a randomized ablation or evidence that larger trades were better",
+      "sourceLocation": "Production Behavior"
+    }
+  ]
+}

source_context/market_behavior.json

@@ -0,0 +1,66 @@
+{
+  "schemaVersion": "dxrg-terminal-pro-market-behavior-observations/v1",
+  "version": "1.0.0",
+  "title": "DX Terminal Pro Market Behavior Observations",
+  "publishedDate": "2026-08-30",
+  "publisher": "DX Research Group",
+  "canonicalPage": "https://www.dxrg.ai/blogs/dx-terminal-pro",
+  "evidenceClass": "historical-live-deployment",
+  "deploymentScope": "One bounded 21-day deployment with a frozen production prompt and harness, one model family, and a 12-token onchain market.",
+  "primarySource": {
+    "title": "Operating-Layer Controls for Onchain Language-Model Agents Under Real Capital",
+    "frozenVersion": "arXiv:2604.26091v1",
+    "url": "https://arxiv.org/abs/2604.26091v1",
+    "localPath": "/blogs/5/dx-terminal-pro-paper.pdf"
+  },
+  "definitions": {
+    "sellCascade": "At least 10 vaults selling the same token within 10 minutes.",
+    "twoSidedTokenWindow": "A five-minute window containing at least one buy and at least one sell for the same token."
+  },
+  "observations": [
+    {
+      "observationId": "feet-one-hour-buy-cascade",
+      "measurement": "buy-cascade",
+      "eventDay": 3,
+      "tokenLabel": "FEET",
+      "windowMinutes": 60,
+      "buyingVaults": 1544,
+      "activeVaults": 3454,
+      "communicationState": "The vaults did not communicate with each other.",
+      "sourceLocator": "Production Behavior Under a Frozen Harness, page 9",
+      "interpretation": "Shared market state was sufficient for a concentrated buy event in the reported deployment."
+    },
+    {
+      "observationId": "poopcoin-largest-sell-cascade",
+      "measurement": "largest-reported-sell-cascade",
+      "tokenLabel": "POOPCOIN",
+      "sellCount": 438,
+      "medianInterAgentGapSeconds": 9.5,
+      "sourceLocator": "Production Behavior Under a Frozen Harness, page 10",
+      "interpretation": "The row describes the largest sell cascade reported for this token in the deployment."
+    },
+    {
+      "observationId": "tournament-sell-cascade-count",
+      "measurement": "sell-cascade-count",
+      "cascadeCount": 3878,
+      "minimumSellingVaults": 10,
+      "sameTokenRequired": true,
+      "windowMinutes": 10,
+      "sourceLocator": "Production Behavior Under a Frozen Harness, page 10",
+      "interpretation": "The count uses the paper's registered sell-cascade definition across the tournament."
+    },
+    {
+      "observationId": "two-sided-token-window-share",
+      "measurement": "trade-share-in-two-sided-token-windows",
+      "tradeSharePercent": 92.9,
+      "windowMinutes": 5,
+      "minimumBuys": 1,
+      "minimumSells": 1,
+      "sameTokenRequired": true,
+      "sourceLocator": "Production Behavior Under a Frozen Harness, page 10",
+      "interpretation": "The shared model produced both buy and sell flow when mandates, positions, and settings differed."
+    }
+  ],
+  "privacyBoundary": "The object contains aggregate event and market-window measurements already published in the paper. It excludes wallet, vault, and agent identifiers, strategy or chat text, reasoning traces, individual trades, individual returns, and language-cohort membership.",
+  "evidenceBoundary": "These rows describe one historical deployment. They do not isolate a single cause or establish ordinary-market prevalence, alpha, profitability, universal safety, or transfer beyond the reported setting."
+}