korrents

On the map

Tap a claim on the ring to put it at the centre.

← When a format reward is simply added to the correctness reward, a…

17 connected korrents · 15 moments from 1 Oct 2023 to 26 Oct 2026.

Everything filed under reinforcement learning reinforcement learning Everything filed under LLMs LLMs Everything filed under AI alignment AI alignment Everything filed under robotics robotics Everything filed under scaling laws scaling laws Everything filed under writing writing Same subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subject Read this korrent: When a format reward is simply added to the correctness reward, a model may over-emphasize the easy format reward and neglect correctness. When a format reward is simply added tothe correctness reward, a model mayover-emphasize the easy format reward andneglect correctness. Last stated a month ago 10 Oct 2026 SR Sebastian Raschka — holds since 2026-10-10 — tap for who they are Same subject: A format reward should only be added when the answer is already correct. — tap to centre the map on it A format reward should only be addedwhen the answer is already correct. Last stated a month ago 10 Oct 2026 SR Sebastian Raschka — holds since 2026-10-10 — tap for who they are Same subject: Editing your writing with a model pulls it towards one homogeneous style, far enough that you lose track of what your own voice was. — tap to centre the map on it Editing your writing with a modelpulls it towards one homogeneousstyle, far enough that you losetrack of what your own voice was. Last stated 3 months ago 19 Aug 2026 AO Addy Osmani — holds since 2026-08-19 — tap for who they are Same subject: When you want a model to work with specific material, pasting the material in works better than relying on its recall. — tap to centre the map on it When you want a model to work withspecific material, pasting thematerial in works better thanrelying on its recall. Last stated 2 years ago 5 Feb 2025 AK Andrej Karpathy — holds since 2025-02-05 — tap for who they are Same subject: Process reward models, such as the additional grader in DeepSeekMath V2, are an interesting direction. — tap to centre the map on it Process reward models, such as theadditional grader in DeepSeekMathV2, are an interesting direction. Last stated a month ago 17 Oct 2026 SR Sebastian Raschka — holds since 2026-10-17 — tap for who they are Same subject: Chain-of-thought explanations from language models can be biased because training does not explicitly reward faithful reasoning. — tap to centre the map on it Chain-of-thought explanations fromlanguage models can be biasedbecause training does not explicitlyreward faithful reasoning. Last stated 2 years ago 1 May 2025 LW Lilian Weng — holds since 2025-05-01 — tap for who they are Same subject: The more uncertain what follows a token, the more useful it is for a model to represent all relevant context on it. — tap to centre the map on it The more uncertain what follows atoken, the more useful it is for amodel to represent all relevantcontext on it. Last stated 11 months ago 20 Dec 2025 NN Neel Nanda — holds since 2025-12-20 — tap for who they are Same subject: The special reasoning-trace delimiter tokens used by reasoning models are cosmetic and not required for good reasoning performance. — tap to centre the map on it The special reasoning-tracedelimiter tokens used by reasoningmodels are cosmetic and not requiredfor good reasoning performance. Last stated 4 months ago 18 Jul 2026 SR Sebastian Raschka — holds since 2026-07-18 — tap for who they are Same subject: Newer reasoning models produce 'wait'-style self-corrections less often than early ones. — tap to centre the map on it Newer reasoning models produce'wait'-style self-corrections lessoften than early ones. Last stated a month ago 17 Oct 2026 SR Sebastian Raschka — holds since 2026-10-17 — tap for who they are Same subject: "Scaling" was powerful because it was one word: naming a research direction is what tells a whole field what to do next. — tap to centre the map on it "Scaling" was powerful because itwas one word: naming a researchdirection is what tells a wholefield what to do next. Last stated 12 months ago 25 Nov 2025 IS Ilya Sutskever — holds since 2025-11-25 — tap for who they are Same subject: A benchmark result should be reported under a stated budget, or as a curve against test-time compute — never as a single number. — tap to centre the map on it A benchmark result should bereported under a stated budget, oras a curve against test-time compute— never as a single number. Last stated 5 months ago 26 Jun 2026 NB Noam Brown — holds since 2026-06-26 — tap for who they are Same subject: A CPU and memory crunch is coming. — tap to centre the map on it A CPU and memory crunch is coming. Last stated 4 weeks ago 26 Oct 2026 DM Dillon Mulroy — holds since 2026-10-26 — tap for who they are Same subject: "Hire smart people and get out of their way" is bad management advice; leaders should stay closely involved, building trust and working through hard problems with their people to create alignment. — tap to centre the map on it "Hire smart people and get out oftheir way" is bad management advice;leaders should stay closelyinvolved, building trust and workingthrough hard problems with theirpeople to create alignment. Last stated a month ago 9 Oct 2026 CR Carl Rivera — holds since 2026-10-09 — tap for who they are Same subject: A 5% level of AI risk is crazy and unacceptable; we would not accept a 5% chance of massive fatalities from paint companies. — tap to centre the map on it A 5% level of AI risk is crazy andunacceptable; we would not accept a5% chance of massive fatalities frompaint companies. Last stated a year ago 8 Jul 2025 RG Ryan Greenblatt — holds since 2025-07-08 — tap for who they are Same subject: A big misreading of Chinese labs is missing that they have a risk posture and do not want to cause harm. — tap to centre the map on it A big misreading of Chinese labs ismissing that they have a riskposture and do not want to causeharm. Last stated 4 weeks ago 26 Oct 2026 NL Nathan Lambert — holds since 2026-10-26 — tap for who they are Same subject: A highly dynamic economy that cheaply mass-produces robots may increase rather than decrease the risk of Malthusian resource dilemmas. — tap to centre the map on it A highly dynamic economy thatcheaply mass-produces robots mayincrease rather than decrease therisk of Malthusian resourcedilemmas. Last stated 3 years ago 1 Oct 2023 TC Tyler Cowen — holds since 2023-10-01 — tap for who they are Same subject: A humanoid robot is the wrong tool on a factory floor, because a machine specialised in the task beats it. — tap to centre the map on it A humanoid robot is the wrong toolon a factory floor, because amachine specialised in the taskbeats it. Last stated 2 years ago 6 Apr 2025 MH Molson Hart — holds since 2025-04-06 — tap for who they are Same subject: A physical AI agent has to be safe on day one, because you cannot ship something merely good enough and let users find the edge cases. — tap to centre the map on it A physical AI agent has to be safeon day one, because you cannot shipsomething merely good enough and letusers find the edge cases. Last stated 4 months ago 3 Aug 2026 DD Dmitri Dolgov — holds since 2026-08-03 — tap for who they are
same subject or similar wordinga cloud: claims about one subject, named for itbar: when it was last stated, on a scale from 2015 to today — full is todaya face: someone who holds the claim — tap it for who they are

At the centre When a format reward is simply added to the correctness reward, a model may over-emphasize the easy format reward and neglect correctness. Last stated 10 Oct 2026 · a month ago Holds Sebastian Raschka Read this korrent →