external_results: ValorBrain BEAM-100K (80.8% / 70.9%) - #36
Open
agentxagi wants to merge 2 commits into
Open
Conversation
Two runs on the beam/100k split (400 queries, 20 conversations): - ValorBrain: 80.8% (323/400) — stealth/ox-alpha reader at reasoning.effort=max, glm-5.2 judge. First run above 80% on this harness to our knowledge. - ValorBrain (GLM-5.2 reader): 70.9% (312/400) — glm-5.2 reader, Gemini 3.6 Flash judge. Both reproducible via the ValorBrain provider in ValorBrain/valorbrain-amb (public API, same model as Hindsight/mem0-cloud). Full writeups linked in each entry's source_url.
|
Someone is attempting to deploy a commit to the Vectorize Team on Vercel. A member of the Team first needs to authorize it. |
200 queries, 10 giant documents (~110M tokens total). stealth/ox-alpha reader at reasoning.effort=max, glm-5.2 judge — same config as our BEAM-100K 80.8% run. Runner accuracy 51.2% (112/200 binary correct), +10.6pp over Hindsight RAG (40.6%) on this split. Full run artifact: results/beam-10m-ox-alpha-effortmax-200q.json in ValorBrain/valorbrain-amb.
Author
|
Model identity correction: the Notable: the reader and judge are both Z.ai models (GLM-5.3-flash reader, glm-5.2 judge). The judge scores against gold answers with rubrics, and the same configuration was used across all runs (including the 70.9% GLM-5.2-reader baseline), so comparisons remain internally consistent. A flash-tier open-weights model beating Gemini-based readers at this price class is the headline. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Adds two ValorBrain runs to the beam/100k section of
external_results.json:Reproducibility
Same model as Hindsight/mem0-cloud: proprietary engine, public API, reproducible results.
ValorBrain/valorbrain-amb—--memory valorbrainresults/beam-100k-ox-alpha-effortmax-400q.jsonin that repoNotes
reasoning.effort: max; retrieval is dense halfvec HNSW + BM25 hybrid inside PostgreSQL.Happy to adjust format/labels to match how you ingest external entries.