{"version":"1.1.0","sourceCommit":"5091f70608a9305cb7bbbf5bb292d60c4c9b361e","principle":"These artifacts report bounded experiments. They do not establish general performance, and campaign failures remain part of the result.","evaluations":[{"id":"planning-gap-2026-09-30","kind":"benchmark","measuredAt":"2026-09-30","command":"node backend/tests/search/test_planning_gap.js","budget":240,"tasks":12,"rawArtifact":"/benchmarks/planning-gap/2026-09-30-a-star-240.json","summary":{"react":{"success":9,"total":12,"avgExpansions":9.2},"tot":{"success":11,"total":12,"avgExpansions":30.6},"mcts":{"success":6,"total":12,"avgExpansions":175.9},"genos":{"success":12,"total":12,"avgExpansions":24.8}},"corrections":{"genosPolicy":"A* with admissible domain heuristics and best known path cost per state","trapFarKey":"moved k3 to reachable cell 2,6; prior key at 6,6 was inaccessible behind its own locked door"}},{"id":"campaign-2026-09-25T14-31-31-114Z","kind":"recorded-campaign","provenance":{"runId":"campaign-2026-09-25T14-31-31-114Z","sourceCommit":"b07e5426c93a3da506f79a5c9c581c99ddedac6b","sourceArtifact":"artifacts/topology-morphogenesis/campaign-2026-09-25T14-31-31-114Z/campaign-results.json","executor":"local","model":"qwen2.5:14b","database":"isolated campaign database","qualification":"historical experimental campaign; verification failed","capturedAt":"2026-09-25","sha256":"01D6434C6608D923C47808BCD791AC793CEB10C1B69715222C36A2F40E8251B3"},"summary":{"missions":12,"verifiedMissions":0,"acceptedDispatches":2,"timeouts":3,"totalWorkerErrors":8,"sessionProbeVerified":1},"missions":[{"name":"orchestrateur-simple","exitCode":1,"timedOut":false,"durationMs":306829,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["orchestrator receipt missing","process exited with code 1","completion gate did not authorize the mission","independent arithmetic proof missing or invalid"],"sessionProbeVerified":null},{"name":"topologie-trinity","exitCode":1,"timedOut":false,"durationMs":161854,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["orchestrator receipt missing","process exited with code 1","no persisted topology workers"],"sessionProbeVerified":null},{"name":"topologie-a-team","exitCode":1,"timedOut":false,"durationMs":146682,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["orchestrator receipt missing","process exited with code 1","no persisted topology workers"],"sessionProbeVerified":null},{"name":"topologie-biome","exitCode":0,"timedOut":false,"durationMs":276423,"dispatchStatus":"accepted","workerStatuses":["error","error","error","error"],"passed":false,"failures":["4 topology worker(s) are not completed"],"sessionProbeVerified":true},{"name":"topologie-biocenose","exitCode":0,"timedOut":false,"durationMs":268321,"dispatchStatus":"accepted","workerStatuses":["error","error","error","error"],"passed":false,"failures":["4 topology worker(s) are not completed"],"sessionProbeVerified":null},{"name":"topologie-holobionte","exitCode":1,"timedOut":false,"durationMs":27768,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["orchestrator receipt missing","process exited with code 1","no persisted topology workers"],"sessionProbeVerified":null},{"name":"topologie-syncytium","exitCode":1,"timedOut":false,"durationMs":15695,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["orchestrator receipt missing","process exited with code 1","no persisted topology workers"],"sessionProbeVerified":null},{"name":"topologie-rhizome","exitCode":null,"timedOut":true,"durationMs":360344,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["mission timed out","orchestrator receipt missing","process exited with code unknown","no persisted topology workers"],"sessionProbeVerified":null},{"name":"topologie-metapopulation","exitCode":1,"timedOut":false,"durationMs":134272,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["orchestrator receipt missing","process exited with code 1","no persisted topology workers"],"sessionProbeVerified":null},{"name":"morphogenese-plan","exitCode":1,"timedOut":false,"durationMs":122577,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["orchestrator receipt missing","process exited with code 1","morphogenesis plan did not pass the completion gate"],"sessionProbeVerified":null},{"name":"morphogenese-shadow","exitCode":null,"timedOut":true,"durationMs":360027,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["mission timed out","orchestrator receipt missing","process exited with code unknown","V2 shadow decision missing","shadow run committed or commit state is unproven","shadow mission completion gate did not pass"],"sessionProbeVerified":null},{"name":"garde-preuve-negative","exitCode":null,"timedOut":true,"durationMs":360026,"dispatchStatus":null,"workerStatuses":[],"passed":false,"failures":["mission timed out","orchestrator receipt missing","negative control did not prove an evidence-based block"],"sessionProbeVerified":null}],"rawArtifact":"/recorded-runs/topology-campaign.json"},{"id":"biocenose-live-2026-10-04","kind":"recorded-campaign","sourceCommit":"5091f70608a9305cb7bbbf5bb292d60c4c9b361e","measuredAt":"2026-10-04","topology":"biocenose","missionCount":12,"cycleStates":{"blocked":12,"completed":0},"fullySatisfiedMissions":0,"verifiedPromotions":0,"reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-qualite-preuves/missions-live-biocenose-2026-10-04.md"},{"id":"biocenose-sqlite-live-2026-10-04","kind":"recorded-campaign","sourceCommit":"5091f70608a9305cb7bbbf5bb292d60c4c9b361e","measuredAt":"2026-10-04","topology":"biocenose","environment":"Full local backend · SQLite WAL · Ollama qwen2.5-coder:7b shared by all members","missionCount":12,"cycleStates":{"blocked":9,"completed":3},"fullySatisfiedMissions":0,"verifiedPromotions":0,"persistence":{"telemetryEvents":766,"communityEvents":549,"reportedLoss":0},"limitations":["A completed cycle can still end in escalation or require additional evidence.","No mission met every criterion in its complex prompt.","Independent provider fault domains were not measured: every member used the same local model.","The Human–AI variant had no human decision and the Hybrid Oracle variant had no injected deterministic receipt."],"reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-qualite-preuves/missions-live-biocenose-sqlite-2026-10-04.md","rawArtifact":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-qualite-preuves/biocenose-sqlite-live-results-2026-10-04.json"}],"publishedRecords":[{"id":"01","status":"REPORTED EVALUATION","tone":"reported","title":"LoCoMo · conversational memory","intro":"Evaluation of GenOS V3 and Qwen 2.5 Coder 7B on the LoCoMo corpus, with 1,986 questions across five categories.","metrics":[["18.39%","Overall F1 · 1,986 questions"],["23.68%","F1 · factual questions (cats. 1–4)"],["5.84%","Overall exact match"]],"detail":"By category: event 30.71%, factual 17.02%, causal 16.92%, temporal 13.23%, adversarial 0.09%. The adversarial score penalizes explicit refusals when the gold answer is “undefined”; the report proposes a separate abstention metric.","source":"Read the LoCoMo report","href":"docs/06-qualite-preuves/benchmarks/locomo.md","question":"What absolute recall score did this GenOS configuration achieve on LoCoMo? No same-model raw-context ablation was reported.","environment":"GenOS V3 · Qwen 2.5 Coder 7B · LoCoMo corpus (1,986 questions)","reproduction":"See the LoCoMo report for corpus version and harness commit.","reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-qualite-preuves/benchmarks/locomo.md"},{"id":"02","status":"PARTIAL DATASET EVALUATION","tone":"reported","title":"SWE-bench Lite · bug resolution","intro":"Run under WSL Ubuntu 24.04 with DeepSeek Coder V2. The report covers tasks ready in the environment, not the entire benchmark.","metrics":[["9 / 26","Tasks resolved"],["34.6%","Of 26 ready tasks"],["300","Tasks in the full dataset"]],"detail":"Fixes are validated with FAIL_TO_PASS and PASS_TO_PASS tests. The 26-task denominator limits the scope: this rate is not a result on all 300 SWE-bench Lite tasks.","source":"Read the SWE-bench Lite report","href":"docs/06-qualite-preuves/benchmarks/swe-bench-lite.md","question":"Can GenOS resolve real bug-fix tasks end to end?","environment":"WSL Ubuntu 24.04 · DeepSeek Coder V2 · 26 ready tasks of 300","reproduction":"FAIL_TO_PASS + PASS_TO_PASS validation per the linked report.","reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-qualite-preuves/benchmarks/swe-bench-lite.md"},{"id":"03","status":"LOCAL EXPERIMENT · SYNTHETIC","tone":"local","title":"AGOW · diffusion and ablations","intro":"Local campaign with Qwen 2.5 Coder 7B: ablation, controlled mediation, and three paired replications on synthetic cases.","metrics":[["11 / 12","Success · full diffusion"],["0 / 12","Success · diffusion removed"],["8 / 8 × 3","Replications · diffusion transmitted"]],"detail":"Workspace ablation scores 11/12 versus 11/12 in this campaign. The same report also documents a Lipson-like run of a simulated trajectory (reward 7.7369) and MBH-RNN (100,000 iterations, ten simulated trials); neither is a replication or a measure of generalization.","source":"Read the campaign report and its limits","href":"docs/06-benchmarks/campagnes-agow-mbh-lipson.md","question":"Does AGOW diffusion change outcomes on synthetic cases?","environment":"Local campaign · Qwen 2.5 Coder 7B · synthetic cases + 3 paired replications","reproduction":"Campaign report documents seeds and ablation arms.","reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-benchmarks/campagnes-agow-mbh-lipson.md"},{"id":"04","status":"LOCAL SUITE · 12 TASKS","tone":"local","title":"Planning · policy comparison","intro":"A deterministic suite of 12 synthetic tasks compares GenOS with ReAct, Tree of Thoughts, and MCTS, with a budget of 240 expansions.","metrics":[["12 / 12","GenOS · tasks passed"],["11 / 12","Tree of Thoughts"],["24.8","Average expansions · GenOS"]],"detail":"In this run, ReAct passes 9/12 and MCTS 6/12 (175.9 average expansions). This is a small local synthetic suite, not an independent standard benchmark; a scenario correction was also recorded with the result.","source":"View the raw result from September 30","href":"benchmarks/planning-gap/results/2026-09-30-a-star-240.json","question":"Does the GenOS planning policy beat ReAct, ToT, and MCTS at equal budget?","environment":"Deterministic harness · 12 synthetic tasks · 240-expansion budget","reproduction":"benchmarks/planning-gap/results/2026-09-30-a-star-240.json","reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/benchmarks/planning-gap/results/2026-09-30-a-star-240.json"},{"id":"05","status":"DETERMINISTIC ABLATION · 4 SCENARIOS","tone":"local","title":"Self-model ablation","intro":"The harness compares eight arms on the same scenarios to check whether removing layers changes the reference decision chain.","metrics":[["6 / 7","Layers with an observed effect"],["4","Deterministic scenarios"],["0","Decisions changed · memory here"]],"detail":"The result depends on a decision harness without an LLM. Effects from some layers remain to be confirmed in runtime missions; the report explicitly distinguishes behavioral change from a status that is merely displayed.","source":"Read the protocol and detailed verdict","href":"benchmarks/cognitive-key-ablation/README.md","question":"Which self-model layers affect the reference decision chain?","environment":"Decision harness without LLM · 4 deterministic scenarios · 8 arms","reproduction":"benchmarks/cognitive-key-ablation/README.md","reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/benchmarks/cognitive-key-ablation/README.md"},{"id":"06","status":"SIMULATION · EXPLICIT ASSUMPTIONS","tone":"exploratory","title":"Evolution of self mechanisms","intro":"Four simulated environments, three seeds, and 40 generations study which layers survive under hypothetical metabolic costs.","metrics":[["4","Simulated environments"],["3","Seeds per environment"],["40","Generations"]],"detail":"Benefit profiles are defined in the model: the experiment informs selection under these assumptions, not an empirical discovery about the mechanisms. The report also flags an uninterpretable contrast in the predictable environment.","source":"Read the P3 research and its limits","href":"docs/06-benchmarks/recherche-evolutionnaire-soi.md","question":"Which layers survive selection under hypothetical metabolic costs?","environment":"Simulation · 4 environments × 3 seeds × 40 generations","reproduction":"Model assumptions and benefit profiles in the linked research doc.","reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-benchmarks/recherche-evolutionnaire-soi.md"},{"id":"07","status":"STRUCTURAL MEASURE · NO LLM","tone":"exploratory","title":"Cognitive Key · recipe diversity","intro":"Ablation on three problems: the harness compares the structural diversity of cognitive recipes composed at dispatch.","metrics":[["3","Recipes · diversity arm E"],["0.51","Pairwise distance · E"],["1.0","Requirement coverage · E"]],"detail":"Arm D (relevance only) produces one repeated recipe and a distance of 0; E produces three recipes with 1.0 coverage. This result concerns structure, not agent response quality: runtime missions remain to be measured.","source":"Read the protocol, measures, and remaining tasks","href":"benchmarks/cognitive-key-ablation/README.md","question":"Does diversity-arm composition produce structurally varied recipes?","environment":"Ablation harness · 3 problems · no LLM in the loop","reproduction":"benchmarks/cognitive-key-ablation/README.md","reportUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/benchmarks/cognitive-key-ablation/README.md"}],"protocolsWithoutCampaign":[{"name":"A-Team","detail":"Paired protocol defined; a campaign connected to real workers has not run yet.","sourceUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-benchmarks/benchmark-ateam.md"},{"name":"Longitudinal Holobiont","detail":"Twelve-arm protocol; no real campaign has run.","sourceUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-benchmarks/benchmark-longitudinal-holobionte.md"},{"name":"Syncytium","detail":"Aggregator and metrics are described; campaign counters remain to be collected.","sourceUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-benchmarks/benchmark-syncytium.md"},{"name":"EAB · epistemic abstention","detail":"The runner awaits the local LoCoMo corpus and complete predictions; no complete EAB report is provided here.","sourceUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/benchmarks/eab/README.md"},{"name":"SWE-bench Lite · extension","detail":"The reported result covers 26 ready tasks, not the full set of 300.","sourceUrl":"https://github.com/PISSARAW/GenOS/blob/5091f70608a9305cb7bbbf5bb292d60c4c9b361e/docs/06-qualite-preuves/benchmarks/swe-bench-lite.md"}]}