Skip to content

docs(research): uncertainty-aware retrieval analysis (#84) - #137

Closed
jphein wants to merge 2 commits into
mainfrom
docs/uncertainty-research-84
Closed

docs(research): uncertainty-aware retrieval analysis (#84)#137
jphein wants to merge 2 commits into
mainfrom
docs/uncertainty-research-84

Conversation

@jphein

@jphein jphein commented May 23, 2026

Copy link
Copy Markdown
Collaborator

Summary

Closes #84.

Analysis of the MIT CSAIL paper Beyond Binary Rewards: Training LMs
to Reason About Their Uncertainty

(Damani et al., ICLR 2026) for applicability to mempalace's hybrid
search pipeline.

Conclusion: defer the headline technique, ship two transferable kernels.

RLCR (Reinforcement Learning with Calibration Rewards) is an RL
fine-tuning method for generative reasoning LMs. mempalace's
retrieval pipeline is not generative-LM-mediated, so the headline
technique does not transfer directly. The press-release framing
"teach AI to say I'm not sure" collapses two distinct ideas —
model-generation uncertainty (what the paper addresses) and
retrieval-ranking uncertainty (what issue #84 actually asks about).

Two transferable kernels worth a 1-week follow-up spike:

  1. Calibrated confidence field on search hits. Apply isotonic
    regression on top of the existing similarity score using the
    n=200 git-derived probe set. Storage ~1KB, query cost ~µs,
    zero new dependencies, no model training. Surface as an optional
    field (not a suppression gate, per the fork's
    closets-as-signal-not-gate philosophy).
  2. Brier-score column in scripts/eval_multi_encoder_rrf.py
    so every future retrieval change is evaluated on calibration
    alongside MRR and Recall. ~20 LOC, permanent value-add even if
    the calibration kernel itself never ships.

The doc covers paper summary with verbatim formulas + Qwen2.5-7B
benchmark table; per-source calibration treatment for
vector/BM25/AGE-graph candidates; cost-vs-fidelity tradeoffs across
three label sources; explicit non-transfer list (RL training of
embeddings, LLM-verbalised confidence per hit, confidence-weighted
majority vote); revisit conditions (RAG endpoint shipping,
drawer-read telemetry available, hybrid becoming the default
candidate strategy).

Files

  • docs/research/uncertainty-aware-retrieval.md (new, ~21KB)
  • docs/fork-changes.yaml — new `uncertainty-aware-retrieval-research` entry
  • FORK_CHANGELOG.md, README.md — re-rendered from yaml

Test plan

  • Skim the analysis — does the "RLCR doesn't transfer; calibration kernel does" framing match the maintainer's read of the paper?
  • Check the per-source calibration discussion (§5) against current _hybrid_rank behaviour
  • Confirm the recommended next-step (1-week spike) shape is the right size before the spike PR lands

No code changes in this PR — analysis only.

Copilot AI review requested due to automatic review settings May 23, 2026 00:54
@gemini-code-assist

Copy link
Copy Markdown

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

@jphein jphein mentioned this pull request May 23, 2026
1 task
@jphein
jphein force-pushed the docs/uncertainty-research-84 branch from 9ea1483 to 90cb6d2 Compare May 23, 2026 01:05
@jphein jphein mentioned this pull request May 23, 2026
1 task
jphein added 2 commits May 22, 2026 18:10
Analyses the MIT CSAIL paper "Beyond Binary Rewards: Training LMs to
Reason About Their Uncertainty" (Damani et al., arXiv:2507.16806,
ICLR 2026) for applicability to mempalace's hybrid search pipeline.

Conclusion: RLCR (the headline technique) does not transfer directly
because it is an RL fine-tuning method for generative reasoning LMs
and mempalace's retrieval pipeline is not generative-LM-mediated.

Two transferable kernels identified for a follow-up spike:

1. Calibrated confidence field on search hits, fit via isotonic
   regression on existing probe sets. Storage ~1KB, query cost
   sub-microsecond, zero new dependencies.
2. Brier-score column in scripts/eval_multi_encoder_rrf.py so every
   future retrieval change is evaluated on calibration alongside MRR
   and Recall.

The doc covers: paper summary with verbatim formulas + Qwen2.5-7B
benchmark table; per-source calibration treatment for vector/BM25/AGE
candidates; why a confidence threshold should not become a suppression
gate (closets-as-signal-not-gate philosophy); cost-vs-fidelity tradeoffs
across three label sources; explicit non-transfer list; revisit
conditions for the headline technique.

Recommendation: defer the headline RLCR integration, spike the
calibration kernel behind a config flag, ship the Brier-score eval
column unconditionally.

Closes #84.
Fills in the previously-TBD commit field on the uncertainty-aware-retrieval-research
fork-changes entry and re-renders FORK_CHANGELOG.md / README.md so the
changelog points at the actual commit instead of a TBD placeholder.
@jphein
jphein force-pushed the docs/uncertainty-research-84 branch from 90cb6d2 to 83d4730 Compare May 23, 2026 01:10
@jphein

jphein commented May 23, 2026

Copy link
Copy Markdown
Collaborator Author

Superseded by clean rebased PR.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

research and integrate

2 participants