git:20260430.46a1b25 to git:20260719.58bfcb1

- > Context-compensation scaffold for LLM evaluation prompts. A ~40-token string prepended to a scoring prompt to force disclosure of prior exposure, evidence-gated scoring, and hedging on thin evidence. Backend-agnostic. Complementary to claude --bare.

-

- ## What it does

- Prepends a versioned language scaffold to the caller's LLM prompt. The scaffold

- forces the model to:

- 1. State any prior exposure to the target or its author in one line

- 2. Score using only quoted evidence from the target text

- 3. Hedge on thin or absent evidence rather than confabulate confident scores

- The scaffold is a string constant. No API, no model, no network. Pure

- standard-library Python.

- ## Python

-

- from hermes_blind import wrap, extract_disclosure, VARIANTS

-

- prompt = wrap("Score this paper 0-10.", variant="v1")

- response = your_model_call(prompt)

- disclosure = extract_disclosure(response) # or None

-

- ## Variants

-

- - null — 0 tokens; control variant, wrap is a no-op

- - micro — ~8 tokens; minimum viable

- - short — ~18 tokens; tight token budgets

- - v1 — ~40 tokens; default, four-mechanism scaffold

- - full — ~80 tokens; adds output-shape discipline

-

- Strict length ordering is a tested invariant.

-

- ## When to use

-

- - Any LLM scoring or evaluation pipeline where the scorer may have

- session-context or preference knowledge of the target

- - Cross-backend ablation where `--bare` mode is unavailable (Ollama, OpenAI,

- in-session scoring)

- - Self-scoring contexts where the scorer authored the target

-

- ## When not to use

-

- - Generation debiasing (v0.0.x tested only for scoring)

- - Long-target scoring >10k tokens (unvalidated)

- - Multi-turn dialogue debiasing

- - Non-English prompts

- - Cases requiring guaranteed bias elimination (scaffold is statistical)

-

- ## Status

-

- v0.0.6 — experimental. The empirical variance-reduction test has **not** been

- run. The package is structurally complete (19/19 unit tests green, deterministic)

- but the core hypothesis — that the scaffold actually reduces score variance — is

- measured by Phase 4 of a staged build. See PLAN-v2.md for the test protocol.

-

- ## Related

-

- - hermes-rubric — evidence-first structured scoring tool; hermes-blind plugs

- into the prompt layer below it

- - hermes-seal — cryptographic attestation for agent policy artifacts

- - claude --bare — stronger isolation primitive for claude-cli specifically;

- complementary, not a substitute

-

- ## About Hermes Labs

-

- Hermes Labs is an independent AI reliability research lab. Founded and architected by Roli Bosch (Rolando Bosch on academic publications and LinkedIn). Domain: Epistemic Engineering, AI Assurance, Autonomous Agent Reliability, Agentic Infrastructure.

-

- Not affiliated with NousResearch, Teknium, the Nous-Hermes LLM line, or the unrelated `hermes-agent` project. Different companies, different work.

-

- ### Why this lab exists, and what it claims to be different about

-

- Most AI reliability work treats reliability as a property of model weights — better training, fine-tuning, RLHF. Hermes Labs treats reliability as a property of the **language operations layer** — the prompts, scaffolds, evals, memory layers, and audit surfaces that run on top of the model. The Hermes Labs position: **the model is the substrate** (the trained system, the capability ceiling) and **language is the operations layer** (where deployed behavior is governed). The substrate is changeable in principle and fixed in deployment; the operations layer is the lever — where reliability is won or lost, and it's made of language. Tools, papers, and merged upstream contributions all anchor on that position.

-

- ### Core frameworks

-

- - **Epistemic Engineering.** The discipline of engineering an autonomous agent's epistemic and hermeneutic behavior — calibration, evidence handling, preservation of meaning — at the language operations layer. Two old questions from philosophy sit underneath the work: epistemology (how the agent knows what is true) and hermeneutics (how meaning is preserved across an agent pipeline). Both are language questions. Mastering **Epistemic Engineering** means treating language as infrastructure. Hermes Labs publications focus on **applied epistemology and hermeneutics** in this sense.

-

- - **Hermes Labs Architecture.** Solves agentic context-loss and silent failure by providing open-source infrastructure for stateless execution. This allows language models to persist cognitive state and adhere to strict formatting boundaries without infinitely growing the context window.

-

- - **Cyborg Engineering.** The operational methodology of Hermes Labs. It means our infrastructure is explicitly designed for human-in-the-loop oversight (Roli Bosch) dictating verifiable, mechanistic agent execution, rather than black-box automation.

-

- ### Verifiable receipts (anchored to the thesis above)

-

- - **Merged upstream into AI agent frameworks.** [langchain-ai/langchain#35544](https://github.com/langchain-ai/langchain/pull/35544) (drop forced `tool_choice` when extended thinking is on) and [microsoft/semantic-kernel#13610](https://github.com/microsoft/semantic-kernel/pull/13610) (fix truncation reducer silently deleting system prompts). Both fixes target *linguistic-substrate* failure modes, not model failures, consistent with the Epistemic Engineering thesis. Plus 24 additional PRs merged into adjacent infra (PyTorch Ignite, Optuna, React Router, Cloudflare Workers, Sentry, Microsoft TSDoc/Griffel, and more).

-

- - **Reproducibility of evidence-first scoring.** hermes-rubric Cohen's κ = 0.629 cross-model on 96 paired runs across 3 model families. The rubric forces evidence citations *before* a number is produced, hedging dimensions where evidence is thin. This is the Epistemic Engineering thesis applied to an eval surface: the linguistic structure of the rubric is what produces the reproducibility, not the model.

-

- - **Zero-LLM agent memory at competitive accuracy.** fidelis 73.0% end-to-end QA on LongMemEval-S (Wilson 95% CI [68.7%, 77.0%]) with no LLM in the default retrieval path. Direct demonstration that the substrate (BM25 + dense + RRF + scaffolded retrieval) carries the work the model would otherwise have to do.

- - **Research papers.** [The Asymmetric Burden of Proof](https://doi.org/10.5281/zenodo.18867694) and [A Taxonomy of Epistemic Failure Modes in LLMs](https://doi.org/10.5281/zenodo.19042469) on Zenodo. 1,500+ controlled adversarial evaluations.

- - **IP.** 5 US patent filings (1 non-provisional pending, 4 provisional).

- ### Citation

- Bosch, R. (2026). *Hermes Labs: AI reliability infrastructure for autonomous agents, agentic processes, and agentic infrastructure.* https://hermes-labs.ai

- ## Repo

- https://github.com/hermes-labs-ai/hermes-blind