{"schema_version":"onlylabs.public_signal.v1","title":"Cohere Repo: cohere-ai/magikarp","description":"Cohere repo signal with public source context, captured evidence pages, related signals, and data-business radar classification.","url":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a","json_url":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a/signal.json","generated_at":"2026-06-11T03:58:56.152276+00:00","org":{"slug":"cohere","name":"Cohere","category":"frontier-lab","category_label":"Frontier lab","dossier_url":"https://onlylabs.fyi/labs/cohere","dossier_json_url":"https://onlylabs.fyi/labs/cohere/dossier.json"},"related_urls":{"signal":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a","signal_json":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a/signal.json","source":"https://github.com/cohere-ai/magikarp","lab_dossier":"https://onlylabs.fyi/labs/cohere","lab_dossier_json":"https://onlylabs.fyi/labs/cohere/dossier.json","analysis":"https://onlylabs.fyi/analysis/cohere","analysis_json":"https://onlylabs.fyi/analysis/cohere/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/cohere/evidence.json","category":"https://onlylabs.fyi/frontier","category_json":"https://onlylabs.fyi/frontier.json","category_feed":"https://onlylabs.fyi/frontier/feed.xml","category_signals_json":"https://onlylabs.fyi/signals.json","topic":null,"topic_signals_json":null,"topic_feed":null,"data_business":null},"answer_pack":{"answer":"Cohere published cohere-ai/magikarp (Python). This repository signal exposes tooling, eval, infrastructure, or model-adjacent work before it may appear in a launch post. High-signal details: repo cohere-ai/magikarp · language Python. onlylabs links this event to 1 captured evidence page and 6 related repo signals.","signal_desk":"repos","source_context":{"source_url":"https://github.com/cohere-ai/magikarp","source_host":"github.com","occurred_at":"2024-04-29T14:29:13+00:00","first_seen_at":"2026-06-05T05:43:00.174979+00:00","date_source":"source","context":"Python"},"context_markers":[{"label":"Lab","value":"Cohere","source":"signal"},{"label":"Signal desk","value":"repos","source":"signal"},{"label":"Source host","value":"github.com","source":"source"},{"label":"Repository","value":"cohere-ai/magikarp","source":"source"},{"label":"Language","value":"Python","source":"source"},{"label":"Stars","value":"191","source":"traction"},{"label":"Watch term","value":"Model card","source":"model"}],"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/cohere-ai/magikarp"],"related_signals":6,"has_source_url":true,"latest_page_fetched_at":"2026-06-11T03:58:56.152276+00:00"},"data_business":{"matches":false,"lanes":[],"matched_terms":[],"score":null,"reason":null},"agent_handoff":{"signal_json":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a/signal.json","dossier_json":"https://onlylabs.fyi/labs/cohere/dossier.json","analysis_json":"https://onlylabs.fyi/analysis/cohere/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/cohere/evidence.json","topic_signals_json":null,"topic_feed":null,"category_signals_json":"https://onlylabs.fyi/signals.json","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","data_business_lane","evidence_url"],"data_business_relevance":"New repositories can expose organization build priorities early, especially around internal tooling, eval infrastructure, data systems, deployment, and agent workflows.","required_sources":[{"label":"signal_json","url":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a/signal.json","required":true},{"label":"source","url":"https://github.com/cohere-ai/magikarp","required":true},{"label":"dossier_json","url":"https://onlylabs.fyi/labs/cohere/dossier.json","required":true},{"label":"analysis_evidence_json","url":"https://onlylabs.fyi/analysis/cohere/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 Cohere's repo signal \"cohere-ai/magikarp\" for frontier lab strategy."},"semantic_triples":[{"subject":"Cohere","predicate":"published repo","object":"cohere-ai/magikarp","text":"Cohere published repo cohere-ai/magikarp."},{"subject":"cohere-ai/magikarp","predicate":"is classified as","object":"repo signal","text":"cohere-ai/magikarp is classified as repo signal."},{"subject":"cohere-ai/magikarp","predicate":"belongs to","object":"repos desk","text":"cohere-ai/magikarp belongs to repos desk."},{"subject":"cohere-ai/magikarp","predicate":"has context","object":"Python","text":"cohere-ai/magikarp has context Python."},{"subject":"cohere-ai/magikarp","predicate":"has evidence coverage","object":"1 captured evidence page","text":"cohere-ai/magikarp has evidence coverage 1 captured evidence page."},{"subject":"cohere-ai/magikarp","predicate":"has captured page count","object":"1","text":"cohere-ai/magikarp has captured page count 1."},{"subject":"cohere-ai/magikarp","predicate":"has readable page count","object":"1","text":"cohere-ai/magikarp has readable page count 1."},{"subject":"cohere-ai/magikarp","predicate":"has related signal count","object":"6","text":"cohere-ai/magikarp has related signal count 6."},{"subject":"cohere-ai/magikarp","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":"cohere-ai/magikarp 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":"cohere-ai/magikarp","predicate":"has source host","object":"github.com","text":"cohere-ai/magikarp has source host github.com."},{"subject":"cohere-ai/magikarp","predicate":"has lab","object":"Cohere","text":"cohere-ai/magikarp has lab Cohere."},{"subject":"cohere-ai/magikarp","predicate":"has signal desk","object":"repos","text":"cohere-ai/magikarp has signal desk repos."},{"subject":"cohere-ai/magikarp","predicate":"has source host","object":"github.com","text":"cohere-ai/magikarp has source host github.com."},{"subject":"cohere-ai/magikarp","predicate":"has repository","object":"cohere-ai/magikarp","text":"cohere-ai/magikarp has repository cohere-ai/magikarp."},{"subject":"cohere-ai/magikarp","predicate":"has language","object":"Python","text":"cohere-ai/magikarp has language Python."},{"subject":"cohere-ai/magikarp","predicate":"has stars","object":"191","text":"cohere-ai/magikarp has stars 191."},{"subject":"cohere-ai/magikarp","predicate":"has watch term","object":"Model card","text":"cohere-ai/magikarp has watch term Model card."}]},"intelligence":{"signal_desk":"repos","answer":"Cohere published cohere-ai/magikarp (Python). This repository signal exposes tooling, eval, infrastructure, or model-adjacent work before it may appear in a launch post. High-signal details: repo cohere-ai/magikarp · language Python. onlylabs links this event to 1 captured evidence page and 6 related repo signals.","semantic_triples":[{"subject":"Cohere","predicate":"published repo","object":"cohere-ai/magikarp","text":"Cohere published repo cohere-ai/magikarp."},{"subject":"cohere-ai/magikarp","predicate":"is classified as","object":"repo signal","text":"cohere-ai/magikarp is classified as repo signal."},{"subject":"cohere-ai/magikarp","predicate":"belongs to","object":"repos desk","text":"cohere-ai/magikarp belongs to repos desk."},{"subject":"cohere-ai/magikarp","predicate":"has context","object":"Python","text":"cohere-ai/magikarp has context Python."},{"subject":"cohere-ai/magikarp","predicate":"has evidence coverage","object":"1 captured evidence page","text":"cohere-ai/magikarp has evidence coverage 1 captured evidence page."}]},"signal":{"id":"db062825-84ba-4cf0-85f1-863d1197844a","url":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a","json_url":"https://onlylabs.fyi/signals/db062825-84ba-4cf0-85f1-863d1197844a/signal.json","source_url":"https://github.com/cohere-ai/magikarp","title":"cohere-ai/magikarp","summary":"Cohere published a new repository. onlylabs watches repos for tooling, eval, infra, and model-adjacent work.","context":"Python","kind":{"key":"repo_new","label":"Repo"},"org":{"slug":"cohere","name":"Cohere","category":"frontier-lab"},"occurred_at":"2024-04-29T14:29:13+00:00","first_seen_at":"2026-06-05T05:43:00.174979+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/cohere-ai/magikarp"]},"facets":{"repo":"cohere-ai/magikarp","language":"Python"},"traction":{"github_stars":191,"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/cohere-ai/magikarp","final_url":"https://github.com/cohere-ai/magikarp","title":"cohere-ai/magikarp repository metadata","http_status":200,"content_type":"application/json","capture_method":"plain","fetched_at":"2026-06-11T03:58:56.152276+00:00","bytes":10804,"raw_path":"20a5ce688e46674c120a541d68fec65cc5629cbf09b333ddcecbac76e98e808b.json","content_hash":"43cfb2bbaa0528d2a0fdca65f721b67fce2c46bccc594aec1846fb3f2cde245c","excerpt_chars":1200,"truncated":true,"excerpt":"cohere-ai/magikarp Description: Code for the paper \"Fishing for Magikarp\" Language: Python License: Apache-2.0 Stars: 191 Forks: 16 Open issues: 2 Created: 2024-04-29T14:29:13Z Pushed: 2026-06-03T23:59:39Z Default branch: main Fork: no Archived: no README: Code for the paper \"Fishing for Magikarp\" This repository contains the code and extended results for the paper Fishing for Magikarp: Automatically Detecting Under-trained Tokens in Large Language Models. The paper is available on [arXiV](https://arxiv.org/abs/2405.05417) and [ACL Anthology](https://aclanthology.org/2024.emnlp-main.649/). Exploring Results The most interesting thing in this repository is probably the detailed reports and summary table found in [results/](results/summary.md). For each model, there is a 'full' and 'mini' report. The 'mini' version can always be opened on github, but the full version may require downloading and viewing locally due to file size limitations. In these reports: * `▁` is a space (but not `_`) * `¿entry?` represents tokens with a vocabulary `entry` which was not encoded as expected. Running on other models Setup
This is a standard This is a standard