Repository navigation
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7eba7b72a6
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f83a9d0ea9
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e58b63f1cb
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review |
|
@codex review |
|
Codex Review: Didn't find any major issues. Chef's kiss. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
|
@codex review Please review the current head |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e9a1dffb58
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review Please review the current head |
|
@codex review Please review the current head |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e6b103fdce
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review Final gate: please review current head |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3ba6fe13fa
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review Final exact-head gate: please review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1e9198ea82
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review Please review the current head |
|
Codex Review: Didn't find any major issues. You're on a roll. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
Summary
This stacked PR builds ModelMap's high-resolution 2025 chapter on top of the corrected 2024 corpus.
Major 2025 releases
Technical lineage additions
Adds four canonical technique nodes:
The slice reuses existing ModelMap concepts for sparse MoE, reasoning-oriented reinforcement learning, and test-time compute rather than creating duplicates.
Evidence-backed adoption edges include DeepSeek-R1 -> reasoning RL; Claude 3.7 -> hybrid reasoning; Qwen3 post-trained variants -> hybrid reasoning + reasoning RL; Llama 4 -> sparse MoE; o3/o4-mini/Claude 4/gpt-oss -> agentic tool use; GPT-5.1 Instant -> adaptive reasoning routing; the GPT-5 unified product system -> adaptive reasoning routing; and DeepSeek-V3.2-Exp -> DeepSeek Sparse Attention. Product-level orchestration is modeled separately from model architecture, and variant-specific training or inference details are not promoted to release-wide edges when the source only establishes them for a subset of models/checkpoints.
Gemini 2.5 Pro remains a thinking-model release but is deliberately not labeled with the repository's
hybrid-reasoningtechnique, because the cited sources do not establish an on/off direct-vs-thinking mode for Pro. Qwen3-Max's September event is scoped to the released Max-Instruct variant while Max-Thinking was still under training. GPT-5.2-Codex is modeled as a variant of GPT-5.2 rather than a separate family release. DeepSeek-V3.2-Speciale is explicitly represented as a V3.2 variant.Family continuity
The OpenAI reasoning lineage is represented as one canonical
family:o-series: both 2024 o1-preview and the full December 2024 o1 release are migrated into the same family as o3-mini, o3, and o4-mini rather than maintaining overlapping o1/o-series family nodes. The full o1 release also carries its own publisher and reasoning-RL edges rather than inheriting them implicitly from o1-preview.The late-2025 closure pass extends already-established Claude, DeepSeek, Gemini, GPT, Mistral, and Qwen family histories through December without introducing partially backfilled new-vendor families.
Source and chronology audit
First-party publication/release dates were rechecked against Anthropic, Google DeepMind, OpenAI, Meta, Qwen, Mistral AI, and DeepSeek sources. DeepSeek-R1 is backed by an immutable official repository snapshot at commit
f7fdff6b839fdb5f9cbdc2696a01d46dee548d32, with evidence locators pointing to the relevant README introduction and post-training sections. Late-2025 additions use first-party release/announcement pages, with publisher edges grounded in organization-specific source context.Historical framing
The 2025 timeline captures the shift from reasoning as a specialized model feature toward hybrid reasoning modes, reasoning-driven tool use, open-weight reasoning, sparse long-context attention, and systems that dynamically route between fast and deeper inference.
Validation
CI validates and generates a canonical 2025 year view in addition to the 2023 and 2024 views. Every relationship remains subject to the hardened endpoint + predicate evidence checks from PR #5.
Stack
Base:
atlas-2024-v0.1(PR #6)Head:
atlas-2025-v0.1