Twitxy Saved

Chris πŸ‡¨πŸ‡¦

Verified account This account is verified. Learn more

@llm_wizard

Open Source Model Lover @ NVIDIA AI

Views my own.
Toronto, Ontario youtube.com/@chrisalexiuk Joined June 2021
840 Following 5.3K Followers 8.3K Posts

Latest posts 5

Chris πŸ‡¨πŸ‡¦ Verified account @llm_wizard
> A common idea floating around is that LLMs will absorb harnesses and make them obsolete.

There is no way enough people believe this for it to be "common".
alex zhang Verified account @a1zhang 8h
Harnesses used to be a crutch for weak LLMs. In 2026 and onward, with stronger models, harnesses now offer a mechanism for composing an LLM’s existing and new capabilities to generalize well beyond their training environments.

In our new paper, LM Harnesses are Compositional Generalizers, we explore how various harness abstractions enable LM systems to generalize over task length, task strategies, recursive task depth, and more through composing existing LLM capabilities. These abstractions provide a means for more data-efficient training, especially over long tasks, by keeping trajectories locally in-distribution (LID) β€” i.e., shaping all individual LLM calls to be in-distribution.

For tasks that take a very long time to roll out, this implies that we can even perform surrogate harness training, meaning we train an individual LLM call instead of doing the very long harness rollout, using purely synthetic tasks, with fake tool calls + sub-agents!

A common idea floating around is that LLMs will absorb harnesses and make them obsolete. We disagree, and argue that we haven’t sufficiently explored what harnesses can provide.

The full paper, logs, codebase, and data are available: alphaxiv.org/abs/2611.harne…

w/ @SOURADIPCHAKR18 @lateinteraction
View quoted post on X
7 replies 0 reposts 34 likes 9.8K views View on X
Chris πŸ‡¨πŸ‡¦ Verified account @llm_wizard
Got to have an amazing conversation with one of Germany's best and brightest!

@DadaJudith has written a lot, and participated a lot in the EU AI scene, and it was a blast to pick her brain at a local pub in Berlin!
NVIDIA Verified account @nvidia 1d
AI is more than a better way to search.

Over a drink with NVIDIA’s Chris Alexiuk, @langdock Co-CEO @DadaJudith shared why AI’s biggest impact may be felt in better services, better products, and healthier, more affordable lives.

Watch their full conversation from GTC Berlin: nvda.ws/4fJhKfE
View quoted post on X
1 replies 0 reposts 11 likes 787 views View on X