{"$schema": "https://iptc.org/std/ninjs/ninjs-schema_3.2.json", "uri": "urn:newsml:wireva.mib.news:story:c46d70a8-cbe5-4601-9604-851e3ee8dbf6", "type": "composite", "version": "1", "versioncreated": "2026-10-18T11:00:23.691253Z", "firstcreated": "2026-10-18T11:00:23.691253Z", "pubstatus": "usable", "language": "en", "headline": "New Benchmark Finds AI Agents Struggle With Long-Term Strategic Choices", "description_text": "A new benchmark called Taste-Bench reveals that even the best AI models make poor strategic decisions in long-running tasks, with GPT-5.6 Sol choosing correctly in just under 60 percent of simple binary questions. Longer deliberation does not improve outcomes.", "associations": {"item-0": {"$schema": "https://iptc.org/std/ninjs/ninjs-schema_3.2.json", "uri": "urn:newsml:wireva.mib.news:item:fa034c9e-502f-478d-98e1-12f2cd9d0aa2", "type": "text", "profile": "standard", "version": "1", "versioncreated": "2026-10-18T11:05:15.544016Z", "firstcreated": "2026-10-18T11:00:23.691253Z", "pubstatus": "usable", "urgency": 5, "language": "en", "headline": "New Benchmark Finds AI Agents Struggle With Long-Term Strategic Choices", "byline": "Jana Hartmann", "located": "Germany", "copyrightholder": "Haydamax OÜ", "copyrightnotice": "Wireva / Hochland", "usageterms": "Monitoring item. Headline, short extract and the source link may be shown. Do not republish the full text.", "description_text": "A new benchmark called Taste-Bench reveals that even the best AI models make poor strategic decisions in long-running tasks, with GPT-5.6 Sol choosing correctly in just under 60 percent of simple binary questions. Longer deliberation does not improve outcomes.", "altids": {"mosaic": "fa034c9e-502f-478d-98e1-12f2cd9d0aa2", "story": "c46d70a8-cbe5-4601-9604-851e3ee8dbf6", "slug": "new-benchmark-finds-ai-agents-struggle-with-long-term-strategic-choices"}, "subject": [{"name": "Technology", "rel": "direct", "literal": "technology", "code": "medtop:13000000", "scheme": "http://cv.iptc.org/newscodes/mediatopic/", "uri": "http://cv.iptc.org/newscodes/mediatopic/13000000"}], "place": [{"name": "Germany", "literal": "de", "code": "DE", "scheme": "http://cv.iptc.org/newscodes/countrycodes/"}], "person": [], "organisation": [], "genre": [{"name": "Standard story", "literal": "standard"}], "service": [{"name": "Wireva", "literal": "wireva"}], "body_text": "A new benchmark called Taste-Bench reveals that even the best AI models make poor strategic decisions in long-running tasks, with GPT-5.6 Sol choosing correctly in just under 60 percent of simple binary questions. Longer deliberation does not improve outcomes.", "extra_body_note": "Extract only — see rights.class = link_only.", "associations": {"featuremedia": {"uri": "urn:newsml:wireva.mib.news:media:9124aba0-bfb8-48a7-b6a2-160dd390ad06", "type": "picture", "version": "1", "pubstatus": "usable", "headline": "AI agents have bad taste in long-term decisions and a new benchmark proves it", "description_text": "New Benchmark Finds AI Agents Struggle With Long-Term Strategic Choices", "copyrightholder": "the-decoder.com", "copyrightnotice": "AI agents have bad taste in long-term decisions and a new benchmark proves it", "usageterms": "Picture belongs to the source publication: address and terms are delivered, the file is not.", "altids": {"wireva": "9124aba0-bfb8-48a7-b6a2-160dd390ad06"}, "extra": {"wireva:delivery_mode": "link_only", "wireva:rights": {"licence": "", "licence_url": "", "licence_raw": "RSS source", "licence_verified_at": "2026-10-18T11:05:15.582353Z", "attribution_required": true, "attribution_text": "AI agents have bad taste in long-term decisions and a new benchmark proves it", "redistribution_allowed": false, "commercial_use_allowed": false, "modifications_allowed": false, "share_alike_required": false}, "wireva:provenance": {"creator": "", "source_name": "the-decoder.com", "source_url": "https://the-decoder.com/wp-content/uploads/2026/10/ai-models-taste-bench-nano-banana-pro.jpg", "source_page_url": "https://the-decoder.com/wp-content/uploads/2026/10/ai-models-taste-bench-nano-banana-pro.jpg", "our_page_url": "https://hochland.uno/technology/new-benchmark-finds-ai-agents-struggle-with-long-term-strategic-choices/"}, "wireva:role": "lead"}, "renditions": {"source": {"href": "https://the-decoder.com/wp-content/uploads/2026/10/ai-models-taste-bench-nano-banana-pro.jpg", "mimetype": "image/jpeg"}}}}, "extra": {"wireva:story": {"uri": "urn:newsml:wireva.mib.news:story:c46d70a8-cbe5-4601-9604-851e3ee8dbf6", "slug": "new-benchmark-finds-ai-agents-struggle-with-long-term-strategic-choices-20261018", "status": "developing", "languages": ["en"], "item_count": 1, "is_breaking": false}, "wireva:rights": {"class": "link_only", "redistribution_allowed": false, "commercial_use_allowed": false, "modifications_allowed": true, "share_alike_required": false, "attribution_required": true, "attribution_text": "Wireva / Hochland", "canonical_required": true, "canonical_url": "https://hochland.uno/technology/new-benchmark-finds-ai-agents-struggle-with-long-term-strategic-choices/", "linkback_required": true, "licence": "", "licence_url": "", "licensor": "Haydamax OÜ", "valid_until": null, "excerpt_max_chars": 280, "decided_by_rule": "derivation_failed", "needs_review": true}, "wireva:provenance": {"production_method": "automated", "ai_stages": {"editing": true, "drafting": true, "headline": true, "discovery": true, "classification": true, "image_selection": true, "source_collection": true, "cross_source_comparison": true}, "model": "", "human_reviewed": false, "reviewed_by": "", "reviewed_at": null, "editorial_control": true, "editorial_responsibility": "Haydamax OÜ (Wireva)", "factchecked": false, "verification_status": "single_source", "independent_confirmations": 0, "official_source_used": false, "used_press_release": false, "public_interest": false, "ai_disclosure_required": false, "ai_disclosure_text": "", "independence_check": {"status": "failed", "score": 0.675, "checked_at": "2026-10-18T11:05:15.576666Z"}, "sources": [{"url": "https://the-decoder.com/ai-agents-have-bad-taste-in-long-term-decisions-and-a-new-benchmark-proves-it/", "title": "AI agents have bad taste in long-term decisions and a new benchmark proves it", "publisher": "the-decoder.com", "type": "rss", "role": "primary", "official": false, "retrieved_at": "2026-10-18T11:05:15.553636Z", "published_at": "2026-10-18T07:49:41Z"}]}, "wireva:publication": {"name": "Hochland", "domain": "hochland.uno", "country": "DE", "url": "https://hochland.uno/technology/new-benchmark-finds-ai-agents-struggle-with-long-term-strategic-choices/"}, "wireva:links": {"self": "https://wireva.mib.news/api/v1/items/fa034c9e-502f-478d-98e1-12f2cd9d0aa2", "html": "https://wireva.mib.news/item/fa034c9e-502f-478d-98e1-12f2cd9d0aa2/new-benchmark-finds-ai-agents-struggle-with-long-term-strategic-choices/", "story": "https://wireva.mib.news/story/new-benchmark-finds-ai-agents-struggle-with-long-term-strategic-choices-20261018/"}, "wireva:metrics": {"word_count": 507, "char_count": 3464, "reading_time_seconds": 247}}}}, "extra": {"wireva:languages": ["en"], "wireva:status": "developing", "wireva:item_count": 1}}