| I think "harness" is a thing, the code/binary, and "agent" is a process, an instantiated run of that code/binary with a given LLM/env, etc. So its like, "GTA 6" as the disc vs. the specific game you're in the middle of being chased by cops, harness vs agent. In practice they're intertwined and it becomes hard not to use the terms somewhat interchangeably, but "you ask the agent how the harness works" vs. the other way around, clearly. |
| |
| ▲ | amdahl 21 hours ago | parent [-] | | It's more like in that analogy, the LLM is the gamer, instantiated as agent within a given game. And the virtual world of GTA 6 is actually your codebase/env, the cops chasing are the bugs/angry customers, etc. The harness is providing an accurate/efficient ability for the model to understand/interact with the virtual world, flee the cops, etc. Decomposed at various architectural boundaries per your taste, but that's like loading a skin on the engine. | | |
| ▲ | bodge5000 21 hours ago | parent [-] | | That doesn't seem right, surely the gamer would still be you, since you interact with the agent through the harness. If the player is the LLM, what is the human in this analogy? I guess a harness doesn't necessarily need human input (most do of course, but thats not a technical limitation), but then again neither does a game for the same reasons Regardless though, this is what I mean, we now have 3 definitions for an agent; an instance of a model (which is how I think of it), a model configuration for a given task (from another commenter) and your definition which appears to be somewhere between the two, though it seems we agree with what a harness is. |
|
|
