Finalize MeSSH26 deck after panel review (v3) - #275
Merged
Conversation
Applied a 4-reviewer panel pass (NLP accuracy, SHS clarity, rhetoric, discussant
red-team). Author decisions: nuance the take-home, merge §7 into the close,
keep concepts light with depth in speaker notes.
Honesty / accuracy:
- Disambiguate "reproductible": the protocol/measures are, the individual LLM
answer is not (limits slide + notes).
- Reframe hallucination/grounding/evaluation as design intentions, not measured
results ("évaluation à venir").
- Fix the pipeline diagram: index node is dense + lexical (BM25 needs a lexical
index), consistent with hybrid retrieval.
- Add a sample caveat to the usage slide (beta, ~190 queries, not representative).
- RGPD annex: state that query text can remain re-identifying (filtered before L3).
- Cost line no longer cherry-picks: name OpenAI dev volume beside the Mistral prod
figure; gloss the RGPD "aller plus loin" as a privacy trade-off, not "better".
Framing / flow:
- Nuance the take-home: "Et si HAL offrait aussi l'accès sémantique ?" plus the
governance/liability/scale obstacles, instead of an imperative "intégrer".
- Merge the usages+coûts section into the close; move the activity timeline to the
appendix. Add the SHS mediation stake as an explicit bullet.
- Demote the English "Retrieval-Augmented Generation" to a parenthetical; gloss R2R.
- Embedding caveat (imperfect, model-dependent); eval metrics named (RAGAS/
faithfulness) in the appendix; demo→§2 bridge and Q&A depth in speaker notes.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Finalization pass on the MeSSH26 slides after a 4-reviewer panel (NLP accuracy,
SHS clarity, rhetoric/timing, discussant red-team).
Author decisions: nuance the take-home ("Et si HAL offrait aussi l'accès
sémantique ?" + governance/scale obstacles), merge §7 (usages+coûts) into the
close, keep the concepts slide light with technical depth in speaker notes.
Honesty/accuracy: disambiguate "reproductible" (protocol vs answer); reframe
hallucination/grounding/eval as intentions not measured results; fix pipeline
index (dense+lexical); add usage-sample caveat; note query-text re-identification
before L3; de-cherry-pick costs (OpenAI dev volume beside Mistral prod figure).
Builds clean-room via
make slides.Ticket: none