{"schema_version":"onlylabs.public_signal.v1","title":"Qwen (Alibaba Cloud) Writing: Global-batch load balance almost free lunch to improve your MoE LLM training","description":"Qwen (Alibaba Cloud) writing signal with public source context, captured evidence pages, related signals, and data-business radar classification.","url":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef","json_url":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef/signal.json","generated_at":"2026-06-07T21:16:13.91553+00:00","org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab","category_label":"Frontier lab","dossier_url":"https://onlylabs.fyi/labs/qwen","dossier_json_url":"https://onlylabs.fyi/labs/qwen/dossier.json"},"related_urls":{"signal":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef","signal_json":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef/signal.json","source":"https://qwenlm.github.io/blog/global-load-balance/","lab_dossier":"https://onlylabs.fyi/labs/qwen","lab_dossier_json":"https://onlylabs.fyi/labs/qwen/dossier.json","analysis":"https://onlylabs.fyi/analysis/qwen","analysis_json":"https://onlylabs.fyi/analysis/qwen/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/qwen/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":"https://onlylabs.fyi/topics/talking","topic_signals_json":"https://onlylabs.fyi/topics/talking/signals.json","topic_feed":"https://onlylabs.fyi/topics/talking/feed.xml","data_business":{"radar":"https://onlylabs.fyi/data-radar","radar_json":"https://onlylabs.fyi/data-radar.json","opportunities":"https://onlylabs.fyi/opportunities","opportunities_json":"https://onlylabs.fyi/opportunities.json","lanes":[{"key":"infrastructure","label":"Infrastructure","url":"https://onlylabs.fyi/data-radar/infrastructure","json_url":"https://onlylabs.fyi/data-radar/infrastructure/signals.json"}]}},"answer_pack":{"answer":"Qwen (Alibaba Cloud) published Global-batch load balance almost free lunch to improve your MoE LLM training. This talking signal gives public context for research themes, product direction, policy, or launch framing. High-signal details: Notable training optimization technique · Global-batch load balance almost free lunch to improve your MoE LLM training | Qwen We have a new blog! View this page at qwen.ai . This page will automatically redirect.... onlylabs links this event to 1 captured evidence page and 6 related writing signals. It also maps to Infrastructure in the data-business radar.","signal_desk":"talking","source_context":{"source_url":"https://qwenlm.github.io/blog/global-load-balance/","source_host":"qwenlm.github.io","occurred_at":"2025-01-21T00:00:00.000Z","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"page.visible_date","context":null},"context_markers":[{"label":"Lab","value":"Qwen (Alibaba Cloud)","source":"signal"},{"label":"Signal desk","value":"talking","source":"signal"},{"label":"Source host","value":"qwenlm.github.io","source":"source"},{"label":"Notability","value":"Notable training optimization technique","source":"signal"},{"label":"Radar lane","value":"Infrastructure","source":"radar"},{"label":"Matched term","value":"training","source":"radar"},{"label":"Watch term","value":"Model card","source":"model"},{"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://qwenlm.github.io/blog/global-load-balance/"],"related_signals":6,"has_source_url":true,"latest_page_fetched_at":"2026-06-07T21:16:13.91553+00:00"},"data_business":{"matches":true,"lanes":[{"key":"infrastructure","label":"Infrastructure","url":"https://onlylabs.fyi/data-radar/infrastructure","json_url":"https://onlylabs.fyi/data-radar/infrastructure/signals.json"}],"matched_terms":["training"],"score":13,"reason":"Qwen (Alibaba Cloud) has a writing signal matching infrastructure."},"agent_handoff":{"signal_json":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef/signal.json","dossier_json":"https://onlylabs.fyi/labs/qwen/dossier.json","analysis_json":"https://onlylabs.fyi/analysis/qwen/analysis.json","analysis_evidence_json":"https://onlylabs.fyi/analysis/qwen/evidence.json","topic_signals_json":"https://onlylabs.fyi/topics/talking/signals.json","topic_feed":"https://onlylabs.fyi/topics/talking/feed.xml","category_signals_json":"https://onlylabs.fyi/signals.json","data_radar_json":"https://onlylabs.fyi/data-radar.json","opportunities_json":"https://onlylabs.fyi/opportunities.json"},"analysis_playbook":{"objective":"Turn public writing and discussion into a readable map of research themes, product framing, policy posture, launch narratives, and market attention.","evidence_focus":["post title","source URL","captured page text","HN traction","linked model or paper references","publication date"],"extraction_questions":["Which themes are labs choosing to explain publicly?","Which posts are attracting outside discussion?","Which writing reframes a recent release, model, hiring wave, or policy stance?","Which posts mention data, evals, infrastructure, safety, or deployment workflows?"],"signal_questions":["What public theme, launch framing, or research direction does this writing signal expose?","Which themes are labs choosing to explain publicly?","Which posts are attracting outside discussion?","Which data-business lane explains this signal: Infrastructure?","Do the 6 related writing signals show a repeated pattern?"],"output_fields":["org","theme","public_framing","traction","data_business_lane","evidence_url"],"data_business_relevance":"Public writing supplies the narrative layer over raw signals and helps identify which frontier-lab priorities are becoming externally legible.","required_sources":[{"label":"signal_json","url":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef/signal.json","required":true},{"label":"source","url":"https://qwenlm.github.io/blog/global-load-balance/","required":true},{"label":"dossier_json","url":"https://onlylabs.fyi/labs/qwen/dossier.json","required":true},{"label":"analysis_evidence_json","url":"https://onlylabs.fyi/analysis/qwen/evidence.json","required":true},{"label":"topic_signals_json","url":"https://onlylabs.fyi/topics/talking/signals.json","required":false},{"label":"data_radar_json","url":"https://onlylabs.fyi/data-radar.json","required":true}],"expected_output":["one-paragraph source-grounded interpretation","data-business 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 Qwen (Alibaba Cloud)'s writing signal \"Global-batch load balance almost free lunch to improve your MoE LLM training\" for frontier lab strategy and data-business implications."},"semantic_triples":[{"subject":"Qwen (Alibaba Cloud)","predicate":"published","object":"Global-batch load balance almost free lunch to improve your MoE LLM training","text":"Qwen (Alibaba Cloud) published Global-batch load balance almost free lunch to improve your MoE LLM training."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"is classified as","object":"writing signal","text":"Global-batch load balance almost free lunch to improve your MoE LLM training is classified as writing signal."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"belongs to","object":"talking desk","text":"Global-batch load balance almost free lunch to improve your MoE LLM training belongs to talking desk."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has evidence coverage","object":"1 captured evidence page","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has evidence coverage 1 captured evidence page."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"matches data-business lanes","object":"Infrastructure","text":"Global-batch load balance almost free lunch to improve your MoE LLM training matches data-business lanes Infrastructure."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has captured page count","object":"1","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has captured page count 1."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has readable page count","object":"1","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has readable page count 1."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has related signal count","object":"6","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has related signal count 6."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has analysis playbook objective","object":"Turn public writing and discussion into a readable map of research themes, product framing, policy posture, launch narratives, and market attention.","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has analysis playbook objective Turn public writing and discussion into a readable map of research themes, product framing, policy posture, launch narratives, and market attention.."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has source host","object":"qwenlm.github.io","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has source host qwenlm.github.io."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has lab","object":"Qwen (Alibaba Cloud)","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has lab Qwen (Alibaba Cloud)."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has signal desk","object":"talking","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has signal desk talking."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has source host","object":"qwenlm.github.io","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has source host qwenlm.github.io."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has notability","object":"Notable training optimization technique","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has notability Notable training optimization technique."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has radar lane","object":"Infrastructure","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has radar lane Infrastructure."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has matched term","object":"training","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has matched term training."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has watch term","object":"Model card","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has watch term Model card."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has watch term","object":"Infrastructure","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has watch term Infrastructure."}]},"intelligence":{"signal_desk":"talking","answer":"Qwen (Alibaba Cloud) published Global-batch load balance almost free lunch to improve your MoE LLM training. This talking signal gives public context for research themes, product direction, policy, or launch framing. High-signal details: Notable training optimization technique · Global-batch load balance almost free lunch to improve your MoE LLM training | Qwen We have a new blog! View this page at qwen.ai . This page will automatically redirect.... onlylabs links this event to 1 captured evidence page and 6 related writing signals. It also maps to Infrastructure in the data-business radar.","semantic_triples":[{"subject":"Qwen (Alibaba Cloud)","predicate":"published","object":"Global-batch load balance almost free lunch to improve your MoE LLM training","text":"Qwen (Alibaba Cloud) published Global-batch load balance almost free lunch to improve your MoE LLM training."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"is classified as","object":"writing signal","text":"Global-batch load balance almost free lunch to improve your MoE LLM training is classified as writing signal."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"belongs to","object":"talking desk","text":"Global-batch load balance almost free lunch to improve your MoE LLM training belongs to talking desk."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"has evidence coverage","object":"1 captured evidence page","text":"Global-batch load balance almost free lunch to improve your MoE LLM training has evidence coverage 1 captured evidence page."},{"subject":"Global-batch load balance almost free lunch to improve your MoE LLM training","predicate":"matches data-business lanes","object":"Infrastructure","text":"Global-batch load balance almost free lunch to improve your MoE LLM training matches data-business lanes Infrastructure."}]},"signal":{"id":"d78248f7-25ac-4b27-beb1-82f3a1988cef","url":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef","json_url":"https://onlylabs.fyi/signals/d78248f7-25ac-4b27-beb1-82f3a1988cef/signal.json","source_url":"https://qwenlm.github.io/blog/global-load-balance/","title":"Global-batch load balance almost free lunch to improve your MoE LLM training","summary":"Qwen (Alibaba Cloud) published a writing signal. onlylabs watches public writing for research themes, product direction, and model-launch context.","context":null,"kind":{"key":"post_published","label":"Writing"},"org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab"},"occurred_at":"2025-01-21T00:00:00.000Z","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"page.visible_date","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://qwenlm.github.io/blog/global-load-balance/"]},"facets":{},"traction":{"github_stars":null,"hn_points":null,"hn_comments":null,"hn_story_id":null,"hf_downloads":null,"hf_likes":null},"data_radar":{"lanes":[{"key":"infrastructure","label":"Infrastructure","url":"https://onlylabs.fyi/data-radar/infrastructure"}],"score":13,"matched_terms":["training"],"reason":"Qwen (Alibaba Cloud) has a writing signal matching infrastructure."}},"primary_evidence_page":{"url":"https://qwenlm.github.io/blog/global-load-balance/","final_url":"https://qwenlm.github.io/blog/global-load-balance/","title":"Global-batch load balance almost free lunch to improve your MoE LLM training","http_status":200,"content_type":"text/html; charset=utf-8","capture_method":"plain","fetched_at":"2026-06-07T21:16:13.91553+00:00","bytes":25578,"raw_path":"a6ee4e994d518c81edf2bc9cfdb68d01c3541ad6729261771631351c5a8a7b2c.html","content_hash":"0f32204374dd159d4bb2fa99b3ee7d560b59da0eae0350c9f7efc6e38db6f066","excerpt_chars":1200,"truncated":true,"excerpt":"Global-batch load balance almost free lunch to improve your MoE LLM training | Qwen We have a new blog! View this page at qwen.ai . This page will automatically redirect in 5 seconds. If you are not redirected automatically, please click the button below. Go Now Global-batch load balance almost free lunch to improve your MoE LLM training January 21, 2025 · 4 min · 739 words · Qwen Team | Translations: 简体中文 GITHUB HUGGING FACE MODELSCOPE DISCORD Background # The Mixture-of-Experts (MoEs) architecture has become a popular model-parameter-scale-up technique. Typically, one MoE layer consists of a router (often parameterized as one single Linear layer) and a group of experts (for transformer-based models, each expert is one feedforward layer). Given an input, only a subset of experts will be activated, and then their outputs will be aggregated based on the scores the router assigned. Specifically, $$ \\mathbb{y}=\\sum_{i\\in N_E,g_i\\in\\operatorname{topK}}g_i(\\mathbb{x})E_i(\\mathbb{x}) $$ Load Balancing loss # Load balancing loss is an essential regularization technique in training MoE-based networks, and high-level intuition encourages the balanced activation of all experts. It can be..."},"evidence_pages":[{"url":"https://qwenlm.github.io/blog/global-load-balance/","final_url":"https://qwenlm.github.io/blog/global-load-balance/","title":"Global-batch load balance almost free lunch to improve your MoE LLM training","http_status":200,"content_type":"text/html; charset=utf-8","capture_method":"plain","fetched_at":"2026-06-07T21:16:13.91553+00:00","bytes":25578,"raw_path":"a6ee4e994d518c81edf2bc9cfdb68d01c3541ad6729261771631351c5a8a7b2c.html","content_hash":"0f32204374dd159d4bb2fa99b3ee7d560b59da0eae0350c9f7efc6e38db6f066","excerpt_chars":1200,"truncated":true,"excerpt":"Global-batch load balance almost free lunch to improve your MoE LLM training | Qwen We have a new blog! View this page at qwen.ai . This page will automatically redirect in 5 seconds. If you are not redirected automatically, please click the button below. Go Now Global-batch load balance almost free lunch to improve your MoE LLM training January 21, 2025 · 4 min · 739 words · Qwen Team | Translations: 简体中文 GITHUB HUGGING FACE MODELSCOPE DISCORD Background # The Mixture-of-Experts (MoEs) architecture has become a popular model-parameter-scale-up technique. Typically, one MoE layer consists of a router (often parameterized as one single Linear layer) and a group of experts (for transformer-based models, each expert is one feedforward layer). Given an input, only a subset of experts will be activated, and then their outputs will be aggregated based on the scores the router assigned. Specifically, $$ \\mathbb{y}=\\sum_{i\\in N_E,g_i\\in\\operatorname{topK}}g_i(\\mathbb{x})E_i(\\mathbb{x}) $$ Load Balancing loss # Load balancing loss is an essential regularization technique in training MoE-based networks, and high-level intuition encourages the balanced activation of all experts. It can be..."}],"related_signals":[{"id":"54029233-5b0e-4748-aae1-6013ae3553d0","url":"https://onlylabs.fyi/signals/54029233-5b0e-4748-aae1-6013ae3553d0","source_url":"https://qwenlm.github.io/blog/qwen3guard/","title":"Qwen3Guard: Real-time Safety for Your Token Stream","context":null,"kind":{"key":"post_published","label":"Writing"},"org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab"},"occurred_at":"2025-09-22T20:00:00+00:00","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"rss.item_date"},{"id":"4e64c872-3d4f-4d01-8908-35b65736eb6e","url":"https://onlylabs.fyi/signals/4e64c872-3d4f-4d01-8908-35b65736eb6e","source_url":"https://qwenlm.github.io/blog/qwen-image-edit/","title":"Qwen-Image-Edit: Image Editing with Higher Quality and Efficiency","context":null,"kind":{"key":"post_published","label":"Writing"},"org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab"},"occurred_at":"2025-08-18T17:30:00+00:00","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"rss.item_date"},{"id":"27bd1b4a-28a4-423c-abb0-e5d24251de65","url":"https://onlylabs.fyi/signals/27bd1b4a-28a4-423c-abb0-e5d24251de65","source_url":"https://qwenlm.github.io/blog/qwen-image/","title":"Qwen-Image: Crafting with Native Text Rendering","context":null,"kind":{"key":"post_published","label":"Writing"},"org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab"},"occurred_at":"2025-08-04T14:08:30+00:00","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"rss.item_date"},{"id":"c662ae5e-1dd9-42a6-8297-87bc670d59ef","url":"https://onlylabs.fyi/signals/c662ae5e-1dd9-42a6-8297-87bc670d59ef","source_url":"https://qwenlm.github.io/blog/gspo/","title":"GSPO: Towards Scalable Reinforcement Learning for Language Models","context":null,"kind":{"key":"post_published","label":"Writing"},"org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab"},"occurred_at":"2025-07-27T07:00:00+00:00","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"rss.item_date"},{"id":"22c0d58c-5ce1-4d1f-97df-de5ebd764517","url":"https://onlylabs.fyi/signals/22c0d58c-5ce1-4d1f-97df-de5ebd764517","source_url":"https://qwenlm.github.io/blog/qwen-mt/","title":"Qwen-MT: Where Speed Meets Smart Translation","context":null,"kind":{"key":"post_published","label":"Writing"},"org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab"},"occurred_at":"2025-07-24T14:00:00+00:00","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"rss.item_date"},{"id":"b6e0bdb2-ffd3-4aed-b44b-732a5e0424a5","url":"https://onlylabs.fyi/signals/b6e0bdb2-ffd3-4aed-b44b-732a5e0424a5","source_url":"https://qwenlm.github.io/blog/qwen3-coder/","title":"Qwen3-Coder: Agentic Coding in the World","context":null,"kind":{"key":"post_published","label":"Writing"},"org":{"slug":"qwen","name":"Qwen (Alibaba Cloud)","category":"frontier-lab"},"occurred_at":"2025-07-22T13:00:00+00:00","first_seen_at":"2026-06-05T05:42:59.088452+00:00","date_source":"rss.item_date"}]}