Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GLM-5.1: Towards Long-Horizon Tasks

z.ai

Post reply on HN
Print view View original

101–110 of 285 posts

Re: GLM-5.1: Towards Long-Horizon Tasks

#101

Post by ReptileMan »

disiplus wrote: post #79
stavros wrote: post #65

Earlier quoted context omitted.

My local tennis court's reservation website was broken and I couldn't cancel a reservation, and I asked GLM-5.1 if it can figure out the API. Five minutes later, I check and it had found a /cancel.php URL that accepted an ID but the ID wasn't exposed anywhere, so it found and was exploiting a blind SQL injection vulnerability to find my reservation ID. Overeager, but I was really really impressed.

Yeah it seems they did not align it to much, at least for now. Yesterday it helped me bypass the bot detection on a local marketplace. that i wanted to scrap some listing for my personal alerting system. Al the others failed but glm5.1 found a set of parameters and tweaks how to make my browser in container not be detected.

Model doing what the user wants with high quality is definitely aligned in my book.

Re: GLM-5.1: Towards Long-Horizon Tasks

#102

Post by gunalx »

HumanOstrich wrote: post #60
bigyabai wrote: post #8

It's an okay model. My biggest issue using GLM 5.1 in OpenCode is that it loses coherency over longer contexts. When you crest 128k tokens, there's a high chance that the model will start spouting gibberish until you compact the history. For short-term bugfixing and tweaks though, it does about what I'd expect from Sonnet for a pretty low price.

I wonder if running the compaction in a degraded state produces a subpar summary to continue with.

Indeed it does. Once i see degraded state i revert to last task and run a compact, before starting up again.

Re: GLM-5.1: Towards Long-Horizon Tasks

#103

Post by girvo »

jauntywundrkind wrote: post #18
bigyabai wrote: post #8

It's an okay model. My biggest issue using GLM 5.1 in OpenCode is that it loses coherency over longer contexts. When you crest 128k tokens, there's a high chance that the model will start spouting gibberish until you compact the history. For short-term bugfixing and tweaks though, it does about what I'd expect from Sonnet for a pretty low price.

Chiming in to second this issue. It is wildly frustrating. I suspect that this isn't the model, but something that z.ai is doing with hosting it. At launch I was related to find glm-5.1 was stable even as the context window filled all the way up (~200k). Where-as glm-5, while it could still talk and think, but had forgotten the finer points of tool use to the point where it was making grevious errors as it went (burn…

This doesn’t help you, but GLM-5 stays coherent far longer on Alibaba’s coding plan/infra. You can’t get that coding plan anymore though unfortunately!

Re: GLM-5.1: Towards Long-Horizon Tasks

#104

Post by throwdbaaway »

jauntywundrkind wrote: post #18
bigyabai wrote: post #8

It's an okay model. My biggest issue using GLM 5.1 in OpenCode is that it loses coherency over longer contexts. When you crest 128k tokens, there's a high chance that the model will start spouting gibberish until you compact the history. For short-term bugfixing and tweaks though, it does about what I'd expect from Sonnet for a pretty low price.

Chiming in to second this issue. It is wildly frustrating. I suspect that this isn't the model, but something that z.ai is doing with hosting it. At launch I was related to find glm-5.1 was stable even as the context window filled all the way up (~200k). Where-as glm-5, while it could still talk and think, but had forgotten the finer points of tool use to the point where it was making grevious errors as it went (burn…

https://github.com/THUDM/IndexCache - Might be some expected issue when rolling out this. They don't have enough compute, and have to innovate.

Re: GLM-5.1: Towards Long-Horizon Tasks

#107

Post by dang »

ray__ wrote: post #28
dang wrote: post #7

Earlier quoted context omitted.

These comments are probably either by friends of the OP or perhaps associated with the project somehow, which is against HN's rules but not the kind of attack we're mostly concerned with these days. Old-fashioned voting rings and booster comments aren't existential threats and actually bring up somewhat nostalgic feelings at the moment! Thanks for watching out for the quality of HN...

Would love to read a Tell HN post about the kinds of attacks you are concerned with!

For example, there are rings of accounts posting generated comments, presumably in order to build karma for spammy or (let's be kind) promotional reasons. There are also plenty of spam rings that create tons of accounts and whatnot.

These are different from the submitter-passed-a-link-to-friends kind of upvoting and booster comments, which feel quaint by comparison. In this case people usually don't know they are breaking HN's rules, which is why they don't try to hide it.

Re: GLM-5.1: Towards Long-Horizon Tasks

#109

Post by pbhjpbhj »

QuantumNomad_ wrote: post #92
adrian_b wrote: post #89

Earlier quoted context omitted.

For conversational purposes that may be too slow, but as a coding assistant this should work, especially if many tasks are batched, so that they may progress simultaneously through a single pass over the SSD data.

Three hour coffee break while the LLM prepares scaffolding for the project.

Like computing used to be. When I first compiled a Linux kernel it ran overnight on a Pentium-S. I had little idea what I was doing, probably compiled all the modules by mistake.
Post reply on HN