Operation Prometheus - SE Trajectories
Collection
GitHub PR trajectories extracted by operation-prometheus for SE model training. • 1 item • Updated
before_context stringlengths 78 1.82k | domain stringclasses 19
values | id stringlengths 18 32 | issue_context stringlengths 45 9.79k | language stringclasses 4
values | outcome stringclasses 1
value | patch stringlengths 2.59k 98.2k | pr_number int64 1 142 | quality_score float64 0.58 0.95 | repo stringlengths 10 31 | review_signals listlengths 1 8 ⌀ | source_license stringclasses 3
values | source_urls listlengths 1 8 | task_type stringclasses 5
values | training_use stringclasses 5
values | validation listlengths 1 4 | license stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
PR title: feat(modulators): implement dynamic neuromodulator gain curves
Changed files (20): .github/workflows/qodana_code_quality.yml, .gitignore, Cargo.lock, Cargo.toml, benches/allocations.rs, benches/encoders.rs, qodana.yaml, src/encoder.rs, src/encoders/delta.rs, src/encoders/derivative.rs, src/encoders/latency.rs... | snn | Limen-Neural-axon-encoder-37 | Issue #26: feat(modulators): Implement dynamic neuromodulator gain curves
## Overview
Extend `modulators.rs` to support real-time dynamic modulation of encoding thresholds, sensitivity, and firing rates based on dopamine/acetylcholine level streams.
## Details
- Refactor `modulators.rs` to support dynamic gain adju... | Rust | merged | # Truncated unified diff for training (full raw under datasets/raw/; 115359 bytes, 20 files)
# added: .github/workflows/qodana_code_quality.yml (+40/-0)
# modified: .gitignore (+32/-7)
# modified: Cargo.lock (+110/-66)
# modified: Cargo.toml (+3/-4)
# modified: benches/allocations.rs (+4/-4)
# modified: benches/encoder... | 37 | 0.95 | Limen-Neural/axon-encoder | [
{
"author": "gemini-code-assist[bot]",
"comment": "## Code Review\n\nThis pull request introduces neuromodulation capabilities across various encoders (DeltaEncoder, PopulationEncoder, PredictiveEncoder, RateEncoder, and TemporalEncoder) by adding support for dynamic gain curves that scale threshold, sensit... | MIT OR Apache-2.0 | [
"https://github.com/Limen-Neural/axon-encoder/pull/37",
"https://github.com/Limen-Neural/axon-encoder/issues/26"
] | feature | review-to-patch | [
{
"detail": "* 'cargo test --locked'\n* 'cargo test --features serde --locked'\n* 'cargo clippy --all-features -- -D warnings'\n* 'cargo test --package axon-encoder --lib -- rng::tests'\n* 'cargo test --message-format=json-diagnostic-rendered-ansi --color=always --no-run --package axon-encoder --lib --profile t... | null |
PR title: Fix Codacy integration and standardize encoders
Changed files (27): .codacy.yml, .github/workflows/ci.yml, .github/workflows/coverage.yml, .gitignore, Cargo.lock, Cargo.toml, README.md, examples/ndarray_encoding.rs, examples/phase_encoding.rs, src/encoder.rs, src/encoders/delta.rs, src/encoders/derivative.rs,... | api | Limen-Neural-axon-encoder-41 | - **New Features**
- Added a Derivative-based encoder with per-channel excitatory/inhibitory spike emission and configurable thresholds.
- Expanded optional Serde support for multiple encoders and related configuration/state, including stricter validation on decode.
- **Bug Fixes**
- Improved spike channel indexi... | Rust | merged | diff --git a/.codacy.yml b/.codacy.yml
new file mode 100644
index 00000000..e8871a54
--- /dev/null
+++ b/.codacy.yml
@@ -0,0 +1,6 @@
+---
+exclude_paths:
+ - "target/**"
+ - "benches/**"
+ - "examples/**"
+ - "tests/**"
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 86acfaa3..3aacab70 100644... | 41 | 0.95 | Limen-Neural/axon-encoder | [
{
"author": "gemini-code-assist[bot]",
"comment": "## Code Review\n\nThis pull request refactors the codebase by moving the `DerivativeEncoder` into the `encoders` module, implementing the `Encoder` trait for it, and adding unit tests. It also introduces `serde` serialization and deserialization support for... | MIT OR Apache-2.0 | [
"https://github.com/Limen-Neural/axon-encoder/pull/41"
] | bugfix | review-to-patch | [
{
"detail": "codecov/patch=success, Code Coverage=success, Validate, Test & Doc=success",
"result": "pass",
"type": "ci"
},
{
"detail": "combined_status=success",
"result": "pass",
"type": "ci"
},
{
"detail": "review_apps: Codacy Static Code Analysis=success, Kilo Code Review=suc... | null |
"PR title: sec(rng): replace insecure xorshift PRNG with rand 0.10.2\nChanged files (9): .gitignore,(...TRUNCATED) | security | Limen-Neural-axon-encoder-50 | "Replaces the custom global `xorshift64*` PRNG (`AtomicU64` + relaxed ordering) in\n`src/rng.rs` wit(...TRUNCATED) | Rust | merged | "diff --git a/.gitignore b/.gitignore\nindex 5d35b2cb..86dff9fd 100644\n--- a/.gitignore\n+++ b/.git(...TRUNCATED) | 50 | 0.95 | Limen-Neural/axon-encoder | [{"author":"gemini-code-assist[bot]","comment":"## Code Review\n\nThis pull request replaces a custo(...TRUNCATED) | MIT OR Apache-2.0 | [
"https://github.com/Limen-Neural/axon-encoder/pull/50"
] | security | repair | [{"detail":"Canonical checklist: **`REVIEW.md`**. Below is what we ran and **what a pass means**.\n\(...TRUNCATED) | null |
"PR title: feat: temporary corpus-ipc decoupling (issues #10, #11, #12, #14)\nChanged files (12): CH(...TRUNCATED) | api | Limen-Neural-brainstem-daemon-24 | "**Temporary decoupling of `corpus-ipc` + `zmq`** so the core library can reach high code quality fi(...TRUNCATED) | Rust | merged | "diff --git a/CHANGELOG.md b/CHANGELOG.md\nindex 17e07e2..24c4bdf 100644\n--- a/CHANGELOG.md\n+++ b/(...TRUNCATED) | 24 | 0.95 | Limen-Neural/brainstem-daemon | [{"author":"gemini-code-assist[bot]","comment":"## Code Review\n\nThis pull request decouples the co(...TRUNCATED) | MIT OR Apache-2.0 | ["https://github.com/Limen-Neural/brainstem-daemon/pull/24","https://github.com/Limen-Neural/brainst(...TRUNCATED) | refactor | review-to-patch | [{"detail":"- `cargo fmt -- --check`\n- `cargo clippy --all-targets --all-features -- -D warnings`\n(...TRUNCATED) | null |
"PR title: Fallible BrainstemDaemon construction with neuron-count validation\nChanged files (3): .g(...TRUNCATED) | snn | Limen-Neural-brainstem-daemon-25 | "- Prevent daemon construction from aborting the process when user-provided config has too many neur(...TRUNCATED) | Rust | merged | "diff --git a/.gitignore b/.gitignore\nindex 3499706..1656ae2 100644\n--- a/.gitignore\n+++ b/.gitig(...TRUNCATED) | 25 | 0.95 | Limen-Neural/brainstem-daemon | [{"author":"gemini-code-assist[bot]","comment":"## Code Review\n\nThis pull request introduces falli(...TRUNCATED) | MIT OR Apache-2.0 | [
"https://github.com/Limen-Neural/brainstem-daemon/pull/25"
] | bugfix | repair | [{"detail":"- Ran `cargo test` and all unit tests passed (`13 passed; 0 failed`).\n- Ran `cargo clip(...TRUNCATED) | null |
"PR title: Migrate daemon to corpus-ipc and neuromod v0.4.0\nChanged files (3): Cargo.toml, README.m(...TRUNCATED) | snn | Limen-Neural-brainstem-daemon-3 | "- switch daemon dependencies from legacy placeholders to `neuromod = 0.4.0` and `corpus-ipc` (with (...TRUNCATED) | Rust | merged | "diff --git a/Cargo.toml b/Cargo.toml\nindex 63da95a..595e4d1 100644\n--- a/Cargo.toml\n+++ b/Cargo.(...TRUNCATED) | 3 | 0.85 | Limen-Neural/brainstem-daemon | [{"author":"chatgpt-codex-connector[bot]","comment":"**<sub><sub> | MIT OR Apache-2.0 | [
"https://github.com/Limen-Neural/brainstem-daemon/pull/3"
] | feature | feature | [{"detail":"- `cargo check --all-targets` in isolated copy (`/tmp/brainstem-daemon-check`) to avoid (...TRUNCATED) | null |
"PR title: feat: resolve open issues #4, #5, #6, #7\nChanged files (14): .codacy.yml, .github/workfl(...TRUNCATED) | infra | Limen-Neural-brainstem-daemon-8 | "Issue #4: Define brainstem-daemon role in runtime/deployment boundary matrix\n\n## Context\n\nThis (...TRUNCATED) | Rust | merged | "# Truncated unified diff for training (full raw under datasets/raw/; 120278 bytes, 14 files)\n# add(...TRUNCATED) | 8 | 0.95 | Limen-Neural/brainstem-daemon | [{"author":"gemini-code-assist[bot]","comment":"## Code Review\n\nThis pull request refactors the `s(...TRUNCATED) | MIT OR Apache-2.0 | ["https://github.com/Limen-Neural/brainstem-daemon/pull/8","https://github.com/Limen-Neural/brainste(...TRUNCATED) | refactor | feature | [{"detail":"validate=success","result":"pass","type":"ci"},{"detail":"combined_status=success","resu(...TRUNCATED) | null |
"PR title: refactor: domain-agnostic neuromodulator API and CI workflow\nChanged files (25): .github(...TRUNCATED) | ml-infra | Limen-Neural-neuromod-15 | "Issue #13: Remove Domain-Specific Mining/HFT Logic from Core SNN Primitive Library\n\n## Context\n\(...TRUNCATED) | Rust | merged | "# Truncated unified diff for training (full raw under datasets/raw/; 196167 bytes, 25 files)\n# add(...TRUNCATED) | 15 | 0.95 | Limen-Neural/neuromod | [{"author":"gemini-code-assist[bot]","comment":"## Code Review\n\nThis pull request refactors the ne(...TRUNCATED) | MIT OR Apache-2.0 | ["https://github.com/Limen-Neural/neuromod/pull/15","https://github.com/Limen-Neural/neuromod/issues(...TRUNCATED) | refactor | review-to-patch | [{"detail":"- [x] `cargo fmt --check`\n- [x] `cargo clippy --all-targets --all-features -- -D warnin(...TRUNCATED) | null |
"PR title: SNN Core: Mining & HFT Purge\nChanged files (9): Cargo.toml, examples/basic.rs, examples/(...TRUNCATED) | snn | Limen-Neural-neuromod-2 | "Refactoring the engine into a pure neuromorphic research library. This strips all blockchain/HFT lo(...TRUNCATED) | Rust | merged | "diff --git a/Cargo.toml b/Cargo.toml\nindex 7dbeaf1..ddaa532 100644\n--- a/Cargo.toml\n+++ b/Cargo.(...TRUNCATED) | 2 | 0.85 | Limen-Neural/neuromod | [{"author":"rmems","comment":"It needs to be remove"},{"author":"rmems","comment":"@copilot apply ch(...TRUNCATED) | MIT OR Apache-2.0 | [
"https://github.com/Limen-Neural/neuromod/pull/2"
] | refactor | review-to-patch | [
{
"detail": "combined_status=pending",
"result": "fail",
"type": "ci"
}
] | null |
"PR title: feat: add Lapicque, Hodgkin-Huxley, FitzHugh-Nagumo, and Hebbian neuron models\nChanged f(...TRUNCATED) | snn | Limen-Neural-neuromod-5 | "Implements the four foundational neuroscience models requested in issue #3, honoring the scientists(...TRUNCATED) | Rust | merged | "diff --git a/README.md b/README.md\nindex 246baaa..382a133 100644\n--- a/README.md\n+++ b/README.md(...TRUNCATED) | 5 | 0.95 | Limen-Neural/neuromod | [{"author":"rmems","comment":"@copilot apply changes based on [this feedback](https://github.com/Spi(...TRUNCATED) | MIT OR Apache-2.0 | ["https://github.com/Limen-Neural/neuromod/pull/5","https://github.com/Limen-Neural/neuromod/issues/(...TRUNCATED) | feature | feature | [{"detail":"Cleanup artifacts=success, Upload results=success, Agent=success, Prepare=success","resu(...TRUNCATED) | null |
End of preview. Expand in Data Studio
Operation Prometheus — software-engineering trajectories
Pull-request engineering trajectories extracted from public GitHub history by operation-prometheus: issue/review signal → code state → patch → validation → outcome.
- Records: 96
- Format: JSONL (
trajectories.jsonl) - Languages: Julia, Python, Rust, SystemVerilog
- sha256:
67090714f80e9482…
Record fields
id, repo, pr_number, source_urls, language, domain, task_type,
before_context, patch, validation, outcome, training_use,
issue_context, review_signals, quality_score, source_license.
training_use buckets:
| Bucket | Records |
|---|---|
bug-prediction |
3 |
feature |
39 |
repair |
22 |
review-to-patch |
21 |
validation |
11 |
Source repositories and license / provenance
Each record's source_license is the license of its originating repository —
this dataset is not under a single blanket license, which is why the card
declares license: other / mixed-per-record rather than MIT or Apache-2.0
for the whole. The export tooling and schemas are Apache-2.0; that does not
relicense source-derived content.
License set represented: Apache-2.0; MIT; MIT OR Apache-2.0.
| Repository | Records | Source license |
|---|---|---|
| Limen-Neural/axon-encoder | 3 | MIT OR Apache-2.0 |
| Limen-Neural/brainstem-daemon | 4 | MIT OR Apache-2.0 |
| Limen-Neural/limbic-critic | 4 | MIT OR Apache-2.0 |
| Limen-Neural/myelin-accelerator | 5 | MIT OR Apache-2.0 |
| Limen-Neural/neuromod | 5 | MIT OR Apache-2.0 |
| Limen-Neural/nir-rs | 4 | MIT OR Apache-2.0 |
| Limen-Neural/synaptic-mesh | 5 | MIT OR Apache-2.0 |
| rmems/LiquidCortex.jl | 3 | MIT OR Apache-2.0 |
| rmems/SpikeStream.jl | 4 | MIT OR Apache-2.0 |
| rmems/Theseus-Quarry | 5 | MIT |
| rmems/agoge-forger | 4 | Apache-2.0 |
| rmems/corinth-canal | 12 | MIT OR Apache-2.0 |
| rmems/engram-parser | 1 | MIT OR Apache-2.0 |
| rmems/grok-ozempic | 12 | MIT OR Apache-2.0 |
| rmems/kinetic-signals | 5 | MIT OR Apache-2.0 |
| rmems/silicon-hdl | 5 | MIT OR Apache-2.0 |
| rmems/spike-viz | 3 | Apache-2.0 |
rmems/vahtisiru (formerly thalamic-relay) |
3 | MIT OR Apache-2.0 |
| rmems/writ | 5 | Apache-2.0 |
| rmems/xai-dissect | 4 | MIT OR Apache-2.0 |
Intended use
Research and evaluation of software-engineering trajectories (feature implementation, repair, validation discipline, review-to-patch). Not cleared for SFT or other training use until an authorship filter has run.
Authorship disclosure
Some source commits carry Co-authored-by: Claude trailers. On each repo's
default branch, rmems/kinetic-signals has 16 of 58 commits and
rmems/xai-dissect has 24 of 73 commits with such a trailer (GitHub API,
2026-10-15). An audit counted 421 such commits across rmems repositories.
Under the project's rule #161 these patches count as hosted-model output.
Gemini and Codex review comments retained as review signals need the same
check. Records have not yet been filtered by authorship.
Contamination and eval overlap
Each row's repo field is matched case-insensitively against the union of two
lists, including their alias entries:
rmems-eval-v0.1-denylist.txt(EXP-014 compose-time train denylist), sha256e785b326f75935a4e709b45e89164c953304baa82af1b794027ab753cd6b2166rmems-eval-v0.1-do-not-train.txt(EXP-008 first-party do-not-train), sha256235dd33b53a8497ca9b031f7503149556c10afc8f3c9428545ef69464109333c
Rows in the denylist's first-party eval freeze section are classed freeze. Rows matching any other entry in either list are classed excluded.
| Class | Repository (repo field) |
Rows | Matched list |
|---|---|---|---|
| freeze | rmems/Theseus-Quarry | 5 | denylist (freeze), do-not-train |
| freeze | rmems/writ | 5 | denylist (freeze), do-not-train (alias) |
| freeze | rmems/xai-dissect | 4 | denylist (freeze), do-not-train |
| excluded | rmems/corinth-canal | 12 | denylist |
| excluded | rmems/grok-ozempic | 12 | denylist |
| excluded | Limen-Neural/myelin-accelerator | 5 | denylist |
| excluded | Limen-Neural/axon-encoder | 3 | denylist |
| excluded | Limen-Neural/neuromod | 5 | do-not-train |
| excluded | Limen-Neural/synaptic-mesh | 5 | do-not-train |
| excluded | rmems/kinetic-signals | 5 | do-not-train |
| excluded | rmems/silicon-hdl | 5 | do-not-train |
| excluded | Limen-Neural/brainstem-daemon | 4 | do-not-train |
| excluded | Limen-Neural/limbic-critic | 4 | do-not-train (alias) |
| excluded | Limen-Neural/nir-rs | 4 | do-not-train |
| excluded | rmems/SpikeStream.jl | 4 | do-not-train |
| excluded | rmems/agoge-forger | 4 | do-not-train |
| excluded | rmems/LiquidCortex.jl | 3 | do-not-train |
| excluded | rmems/spike-viz | 3 | do-not-train |
| excluded | rmems/thalamic-relay | 3 | do-not-train (old name of rmems/vahtisiru) |
| excluded | rmems/engram-parser | 1 | do-not-train |
Totals: freeze 14, excluded 82, eligible 0 (96 rows).
The earlier 9/32/55 count used only the denylist and missed renamed repos. It is superseded by this table.
rmems/myelin-accelerator is listed as the old-owner alias of Limen-Neural/myelin-accelerator (repo owner renamed).
Fail-closed rule:
- All 96 rows are excluded from training until the split is done.
- Freeze rows stay eval-only.
- No row is training-safe until the authorship filter has run.
Stamps:
- Counted revision:
9aa45e30c3de724e91c6760a916742371969324b trajectories.jsonlsha256:67090714f80e94824d3821b4449eae893b3bdc09e86c1d379760b9cfb51142b2
Exclusions and limitations
Per the data policy: no secrets, private configs, model weights, closed-model chat logs, or bot boilerplate (review bots like Gemini/Codex are retained as signal). Large patches may be truncated (~96 KiB cap). Solo-maintainer repos: limited multi-human review. Inspect records before training use.
- Downloads last month
- 149