{"schema_version":"onlylabs.public_signal.v1","title":"AI21 Labs Repo: AI21Labs/multi-window-chunk-size","description":"AI21 Labs repo signal with public source context, captured evidence pages, related signals, and category-scoped analysis context.","url":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f","json_url":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f/signal.json","generated_at":"2026-06-11T02:53:18.112913+00:00","org":{"slug":"ai21","name":"AI21 Labs","category":"neolab","category_label":"Neolab","dossier_url":"https://onlylabs.fyi/labs/ai21","dossier_json_url":"https://onlylabs.fyi/labs/ai21/dossier.json"},"related_urls":{"signal":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f","signal_json":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f/signal.json","source":"https://github.com/AI21Labs/multi-window-chunk-size","lab_dossier":"https://onlylabs.fyi/labs/ai21","lab_dossier_json":"https://onlylabs.fyi/labs/ai21/dossier.json","analysis":"https://onlylabs.fyi/analysis/ai21","analysis_json":"https://onlylabs.fyi/analysis/ai21/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/ai21/evidence.json","category":"https://onlylabs.fyi/neolabs","category_json":"https://onlylabs.fyi/neolabs.json","category_feed":"https://onlylabs.fyi/neolabs/feed.xml","category_signals_json":"https://onlylabs.fyi/signals.json?category=neolab","topic":null,"topic_signals_json":null,"topic_feed":null,"data_business":null},"answer_pack":{"answer":"AI21 Labs published AI21Labs/multi-window-chunk-size (Jupyter Notebook). This repository signal exposes tooling, eval, infrastructure, or model-adjacent work before it may appear in a launch post. High-signal details: repo AI21Labs/multi-window-chunk-size · language Jupyter Notebook · Low-star repo by AI21. onlylabs links this event to 1 captured evidence page and 6 related repo signals.","signal_desk":"repos","source_context":{"source_url":"https://github.com/AI21Labs/multi-window-chunk-size","source_host":"github.com","occurred_at":"2026-01-26T12:02:59+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source","context":"Jupyter Notebook"},"context_markers":[{"label":"Lab","value":"AI21 Labs","source":"signal"},{"label":"Signal desk","value":"repos","source":"signal"},{"label":"Source host","value":"github.com","source":"source"},{"label":"Repository","value":"AI21Labs/multi-window-chunk-size","source":"source"},{"label":"Language","value":"Jupyter Notebook","source":"source"},{"label":"Stars","value":"7","source":"traction"},{"label":"Notability","value":"Low-star repo by AI21","source":"signal"},{"label":"Watch term","value":"Eval methodology","source":"evidence"},{"label":"Watch term","value":"Data pipeline","source":"evidence"},{"label":"Watch term","value":"Infrastructure","source":"evidence"}],"evidence_coverage":{"target_pages":1,"captured_pages":1,"readable_pages":1,"capture_methods":["plain"],"missing_page_urls":[],"failed_page_urls":[],"blocked_page_urls":[],"page_urls":["https://github.com/AI21Labs/multi-window-chunk-size"],"related_signals":6,"has_source_url":true,"latest_page_fetched_at":"2026-06-11T02:53:18.112913+00:00"},"data_business":{"matches":false,"lanes":[],"matched_terms":[],"score":null,"reason":null},"agent_handoff":{"signal_json":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f/signal.json","dossier_json":"https://onlylabs.fyi/labs/ai21/dossier.json","analysis_json":"https://onlylabs.fyi/analysis/ai21/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/ai21/evidence.json","topic_signals_json":null,"topic_feed":null,"category_signals_json":"https://onlylabs.fyi/signals.json?category=neolab","data_radar_json":null,"opportunities_json":null},"analysis_playbook":{"objective":"Turn new repository signals into early evidence of tooling, eval, infrastructure, model-adjacent, or product work before it appears in polished launch channels.","evidence_focus":["repo name","owner","description","language","stars","source URL","first seen time","data, eval, infra, safety, and product terms"],"extraction_questions":["What technical area does this repository expose?","Does the repo imply eval, data, infrastructure, agent, or deployment work?","Is the repo new evidence for a lab direction that is not yet in writing or releases?","Which related signals should an analyst inspect next?"],"signal_questions":["What does this new repository reveal before a formal announcement exists?","What technical area does this repository expose?","Does the repo imply eval, data, infrastructure, agent, or deployment work?","Do the 6 related repo signals show a repeated pattern?"],"output_fields":["org","repo","technical_theme","evidence_url"],"data_business_relevance":"Data-business lane extraction is scoped to frontier labs; for this category, interpret the repository as source-grounded category strategy evidence.","required_sources":[{"label":"signal_json","url":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f/signal.json","required":true},{"label":"source","url":"https://github.com/AI21Labs/multi-window-chunk-size","required":true},{"label":"dossier_json","url":"https://onlylabs.fyi/labs/ai21/dossier.json","required":true},{"label":"analysis_evidence_json","url":"https://onlylabs.fyi/analysis/ai21/evidence.json","required":true},{"label":"topic_signals_json","url":null,"required":false},{"label":"data_radar_json","url":null,"required":false}],"expected_output":["one-paragraph source-grounded interpretation","category-specific implication","confidence and missing evidence","recommended next source to inspect"],"prompt_seed":"Using only the linked onlylabs JSON, captured source context, and cited evidence, analyze AI21 Labs's repo signal \"AI21Labs/multi-window-chunk-size\" for neolab strategy."},"semantic_triples":[{"subject":"AI21 Labs","predicate":"published repo","object":"AI21Labs/multi-window-chunk-size","text":"AI21 Labs published repo AI21Labs/multi-window-chunk-size."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"is classified as","object":"repo signal","text":"AI21Labs/multi-window-chunk-size is classified as repo signal."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"belongs to","object":"repos desk","text":"AI21Labs/multi-window-chunk-size belongs to repos desk."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has context","object":"Jupyter Notebook","text":"AI21Labs/multi-window-chunk-size has context Jupyter Notebook."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has evidence coverage","object":"1 captured evidence page","text":"AI21Labs/multi-window-chunk-size has evidence coverage 1 captured evidence page."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has captured page count","object":"1","text":"AI21Labs/multi-window-chunk-size has captured page count 1."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has readable page count","object":"1","text":"AI21Labs/multi-window-chunk-size has readable page count 1."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has related signal count","object":"6","text":"AI21Labs/multi-window-chunk-size has related signal count 6."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has analysis playbook objective","object":"Turn new repository signals into early evidence of tooling, eval, infrastructure, model-adjacent, or product work before it appears in polished launch channels.","text":"AI21Labs/multi-window-chunk-size has analysis playbook objective Turn new repository signals into early evidence of tooling, eval, infrastructure, model-adjacent, or product work before it appears in polished launch channels.."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has source host","object":"github.com","text":"AI21Labs/multi-window-chunk-size has source host github.com."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has lab","object":"AI21 Labs","text":"AI21Labs/multi-window-chunk-size has lab AI21 Labs."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has signal desk","object":"repos","text":"AI21Labs/multi-window-chunk-size has signal desk repos."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has source host","object":"github.com","text":"AI21Labs/multi-window-chunk-size has source host github.com."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has repository","object":"AI21Labs/multi-window-chunk-size","text":"AI21Labs/multi-window-chunk-size has repository AI21Labs/multi-window-chunk-size."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has language","object":"Jupyter Notebook","text":"AI21Labs/multi-window-chunk-size has language Jupyter Notebook."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has stars","object":"7","text":"AI21Labs/multi-window-chunk-size has stars 7."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has notability","object":"Low-star repo by AI21","text":"AI21Labs/multi-window-chunk-size has notability Low-star repo by AI21."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has watch term","object":"Eval methodology","text":"AI21Labs/multi-window-chunk-size has watch term Eval methodology."}]},"intelligence":{"signal_desk":"repos","answer":"AI21 Labs published AI21Labs/multi-window-chunk-size (Jupyter Notebook). This repository signal exposes tooling, eval, infrastructure, or model-adjacent work before it may appear in a launch post. High-signal details: repo AI21Labs/multi-window-chunk-size · language Jupyter Notebook · Low-star repo by AI21. onlylabs links this event to 1 captured evidence page and 6 related repo signals.","semantic_triples":[{"subject":"AI21 Labs","predicate":"published repo","object":"AI21Labs/multi-window-chunk-size","text":"AI21 Labs published repo AI21Labs/multi-window-chunk-size."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"is classified as","object":"repo signal","text":"AI21Labs/multi-window-chunk-size is classified as repo signal."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"belongs to","object":"repos desk","text":"AI21Labs/multi-window-chunk-size belongs to repos desk."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has context","object":"Jupyter Notebook","text":"AI21Labs/multi-window-chunk-size has context Jupyter Notebook."},{"subject":"AI21Labs/multi-window-chunk-size","predicate":"has evidence coverage","object":"1 captured evidence page","text":"AI21Labs/multi-window-chunk-size has evidence coverage 1 captured evidence page."}]},"signal":{"id":"7beb24f5-17a5-4573-92ef-792037a0120f","url":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f","json_url":"https://onlylabs.fyi/signals/7beb24f5-17a5-4573-92ef-792037a0120f/signal.json","source_url":"https://github.com/AI21Labs/multi-window-chunk-size","title":"AI21Labs/multi-window-chunk-size","summary":"AI21 Labs published a new repository. onlylabs watches repos for tooling, eval, infra, and model-adjacent work.","context":"Jupyter Notebook","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"ai21","name":"AI21 Labs","category":"neolab"},"occurred_at":"2026-01-26T12:02:59+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source","evidence_coverage":{"target_pages":1,"captured_pages":1,"readable_pages":1,"capture_methods":["plain"],"missing_page_urls":[],"failed_page_urls":[],"blocked_page_urls":[],"page_urls":["https://github.com/AI21Labs/multi-window-chunk-size"]},"facets":{"repo":"AI21Labs/multi-window-chunk-size","language":"Jupyter Notebook"},"traction":{"github_stars":7,"hn_points":null,"hn_comments":null,"hn_story_id":null,"hf_downloads":null,"hf_likes":null},"data_radar":null},"primary_evidence_page":{"url":"https://github.com/AI21Labs/multi-window-chunk-size","final_url":"https://github.com/AI21Labs/multi-window-chunk-size","title":"AI21Labs/multi-window-chunk-size repository metadata","http_status":200,"content_type":"application/json","capture_method":"plain","fetched_at":"2026-06-11T02:53:18.112913+00:00","bytes":11309,"raw_path":"062486fa5a6b7f882201520066bcc76b8c57dc247faf89f9db210580f5fc4881.json","content_hash":"4b280b9028c3e8194f9ffe229afd58ff18e0a21750dc0706a5d38797e9adce09","excerpt_chars":1200,"truncated":true,"excerpt":"AI21Labs/multi-window-chunk-size Language: Jupyter Notebook Stars: 7 Forks: 1 Open issues: 0 Created: 2026-01-26T12:02:59Z Pushed: 2026-01-28T12:44:54Z Default branch: master Fork: no Archived: no README: Multi-Scale Retrieval with RRF This repository demonstrates a multi-scale retrieval approach for RAG (Retrieval-Augmented Generation) systems, showing that **chunk size is query-dependent** and that aggregating results across multiple chunk sizes improves retrieval robustness. <p align=\"center\"> <img src=\"assets/multi-scale-retrieval-diagram.png\" alt=\"Multi-Scale Retrieval with RRF Diagram\" width=\"800\"> </p> Overview Instead of committing to a single chunk size, we: 1. Index the same corpus multiple times with different chunk sizes (100, 200, 500 tokens) 2. Query all indices in parallel at inference time 3. Aggregate results using **Reciprocal Rank Fusion (RRF)** to produce final document rankings Repository Structure ``` ├── multi-window-chunk-size.ipynb # Main notebook demonstrating the approach ├── seinfeld_trivia/ │ ├── data.json # Dataset with trivia questions and gold documents │ └── documents_content/ # Markdown files for each Seinfeld episode │ ├── S01E00.md │ ├──..."},"evidence_pages":[{"url":"https://github.com/AI21Labs/multi-window-chunk-size","final_url":"https://github.com/AI21Labs/multi-window-chunk-size","title":"AI21Labs/multi-window-chunk-size repository metadata","http_status":200,"content_type":"application/json","capture_method":"plain","fetched_at":"2026-06-11T02:53:18.112913+00:00","bytes":11309,"raw_path":"062486fa5a6b7f882201520066bcc76b8c57dc247faf89f9db210580f5fc4881.json","content_hash":"4b280b9028c3e8194f9ffe229afd58ff18e0a21750dc0706a5d38797e9adce09","excerpt_chars":1200,"truncated":true,"excerpt":"AI21Labs/multi-window-chunk-size Language: Jupyter Notebook Stars: 7 Forks: 1 Open issues: 0 Created: 2026-01-26T12:02:59Z Pushed: 2026-01-28T12:44:54Z Default branch: master Fork: no Archived: no README: Multi-Scale Retrieval with RRF This repository demonstrates a multi-scale retrieval approach for RAG (Retrieval-Augmented Generation) systems, showing that **chunk size is query-dependent** and that aggregating results across multiple chunk sizes improves retrieval robustness. <p align=\"center\"> <img src=\"assets/multi-scale-retrieval-diagram.png\" alt=\"Multi-Scale Retrieval with RRF Diagram\" width=\"800\"> </p> Overview Instead of committing to a single chunk size, we: 1. Index the same corpus multiple times with different chunk sizes (100, 200, 500 tokens) 2. Query all indices in parallel at inference time 3. Aggregate results using **Reciprocal Rank Fusion (RRF)** to produce final document rankings Repository Structure ``` ├── multi-window-chunk-size.ipynb # Main notebook demonstrating the approach ├── seinfeld_trivia/ │ ├── data.json # Dataset with trivia questions and gold documents │ └── documents_content/ # Markdown files for each Seinfeld episode │ ├── S01E00.md │ ├──..."}],"related_signals":[{"id":"a8d5a050-6c2c-48fc-8704-64a597e69658","url":"https://onlylabs.fyi/signals/a8d5a050-6c2c-48fc-8704-64a597e69658","source_url":"https://github.com/AI21Labs/pre-commit-hadolint","title":"AI21Labs/pre-commit-hadolint","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"ai21","name":"AI21 Labs","category":"neolab"},"occurred_at":"2026-05-12T08:34:02+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source"},{"id":"ac9e3f80-ea99-4dba-8929-da82435a3a34","url":"https://onlylabs.fyi/signals/ac9e3f80-ea99-4dba-8929-da82435a3a34","source_url":"https://github.com/AI21Labs/ai21-typescript","title":"AI21Labs/ai21-typescript","context":"TypeScript","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"ai21","name":"AI21 Labs","category":"neolab"},"occurred_at":"2024-11-05T09:44:25+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source"},{"id":"653c8c77-1a4f-4e50-8504-f5f596fb5be3","url":"https://onlylabs.fyi/signals/653c8c77-1a4f-4e50-8504-f5f596fb5be3","source_url":"https://github.com/AI21Labs/salt","title":"AI21Labs/salt","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"ai21","name":"AI21 Labs","category":"neolab"},"occurred_at":"2024-07-21T09:13:44+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source"},{"id":"d4083267-defa-4b38-8f7f-90ee756c2801","url":"https://onlylabs.fyi/signals/d4083267-defa-4b38-8f7f-90ee756c2801","source_url":"https://github.com/AI21Labs/AI21-Industry-Samples","title":"AI21Labs/AI21-Industry-Samples","context":"Jupyter Notebook","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"ai21","name":"AI21 Labs","category":"neolab"},"occurred_at":"2024-02-02T14:30:39+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source"},{"id":"3ceb6253-41fc-44c7-a33e-62f570e3922b","url":"https://onlylabs.fyi/signals/3ceb6253-41fc-44c7-a33e-62f570e3922b","source_url":"https://github.com/AI21Labs/ai21-python","title":"AI21Labs/ai21-python","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"ai21","name":"AI21 Labs","category":"neolab"},"occurred_at":"2023-12-10T09:20:15+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source"},{"id":"3654e90c-8a76-4a9f-b80e-e7e89c4edb52","url":"https://onlylabs.fyi/signals/3654e90c-8a76-4a9f-b80e-e7e89c4edb52","source_url":"https://github.com/AI21Labs/ai21-tokenizer","title":"AI21Labs/ai21-tokenizer","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"ai21","name":"AI21 Labs","category":"neolab"},"occurred_at":"2023-08-22T08:28:46+00:00","first_seen_at":"2026-06-05T20:58:44.654825+00:00","date_source":"source"}]}